consul/agent/config/runtime.go

1843 lines
63 KiB
Go
Raw Normal View History

New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
package config
import (
"fmt"
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
"net"
"reflect"
"strings"
"time"
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/consul/api"
2018-06-14 12:52:48 +00:00
"github.com/hashicorp/consul/lib"
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
"github.com/hashicorp/consul/tlsutil"
"github.com/hashicorp/consul/types"
"golang.org/x/time/rate"
)
type RuntimeSOAConfig struct {
Refresh uint32 // 3600 by default
Retry uint32 // 600
Expire uint32 // 86400
Minttl uint32 // 0,
}
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
// RuntimeConfig specifies the configuration the consul agent actually
// uses. Is is derived from one or more Config structures which can come
// from files, flags and/or environment variables.
type RuntimeConfig struct {
// non-user configurable values
2017-10-18 15:23:56 +00:00
AEInterval time.Duration
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
CheckDeregisterIntervalMin time.Duration
CheckReapInterval time.Duration
SegmentLimit int
SegmentNameLimit int
SyncCoordinateRateTarget float64
SyncCoordinateIntervalMin time.Duration
Revision string
Version string
VersionPrerelease string
// consul config
ConsulCoordinateUpdateMaxBatches int
ConsulCoordinateUpdateBatchSize int
ConsulCoordinateUpdatePeriod time.Duration
ConsulRaftElectionTimeout time.Duration
ConsulRaftHeartbeatTimeout time.Duration
ConsulRaftLeaderLeaseTimeout time.Duration
ConsulServerHealthInterval time.Duration
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLDisabledTTL is used by agents to determine how long they will
// wait to check again with the servers if they discover ACLs are not
// enabled. (not user configurable)
//
// hcl: acl.disabled_ttl = "duration"
ACLDisabledTTL time.Duration
// ACLsEnabled is used to determine whether ACLs should be enabled
//
// hcl: acl.enabled = boolean
ACLsEnabled bool
2017-10-18 15:23:56 +00:00
// ACLAgentMasterToken is a special token that has full read and write
// privileges for this agent, and can be used to call agent endpoints
// when no servers are available.
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.tokens.agent_master = string
2017-10-18 15:23:56 +00:00
ACLAgentMasterToken string
// ACLAgentToken is the default token used to make requests for the agent
// itself, such as for registering itself with the catalog. If not
// configured, the 'acl_token' will be used.
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.tokens.agent = string
2017-10-18 15:23:56 +00:00
ACLAgentToken string
// ACLDatacenter is the central datacenter that holds authoritative
// ACL records. This must be the same for the entire cluster.
// If this is not set, ACLs are not enabled. Off by default.
//
// hcl: acl_datacenter = string
ACLDatacenter string
// ACLDefaultPolicy is used to control the ACL interaction when
// there is no defined policy. This can be "allow" which means
// ACLs are used to black-list, or "deny" which means ACLs are
// white-lists.
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.default_policy = ("allow"|"deny")
2017-10-18 15:23:56 +00:00
ACLDefaultPolicy string
// ACLDownPolicy is used to control the ACL interaction when we cannot
// reach the ACLDatacenter and the token is not in the cache.
2017-10-19 07:03:17 +00:00
// There are the following modes:
2017-10-18 15:23:56 +00:00
// * allow - Allow all requests
// * deny - Deny all requests
// * extend-cache - Ignore the cache expiration, and allow cached
// ACL's to be used to service requests. This
// is the default. If the ACL is not in the cache,
// this acts like deny.
// * async-cache - Same behavior as extend-cache, but perform ACL
// Lookups asynchronously when cache TTL is expired.
2017-10-18 15:23:56 +00:00
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.down_policy = ("allow"|"deny"|"extend-cache"|"async-cache")
2017-10-18 15:23:56 +00:00
ACLDownPolicy string
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// DEPRECATED (ACL-Legacy-Compat)
2017-10-18 15:23:56 +00:00
// ACLEnforceVersion8 is used to gate a set of ACL policy features that
// are opt-in prior to Consul 0.8 and opt-out in Consul 0.8 and later.
//
// hcl: acl_enforce_version_8 = (true|false)
ACLEnforceVersion8 bool
2017-10-19 07:03:17 +00:00
// ACLEnableKeyListPolicy is used to opt-in to the "list" policy added to
// KV ACLs in Consul 1.0.
//
// See https://www.consul.io/docs/guides/acl.html#list-policy-for-keys for
// more details.
2017-10-18 15:23:56 +00:00
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.enable_key_list_policy = (true|false)
ACLEnableKeyListPolicy bool
2017-10-18 15:23:56 +00:00
// ACLMasterToken is used to bootstrap the ACL system. It should be specified
// on the servers in the ACLDatacenter. When the leader comes online, it ensures
// that the Master token is available. This provides the initial token.
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.tokens.master = string
2017-10-18 15:23:56 +00:00
ACLMasterToken string
ACL Token Persistence and Reloading (#5328) This PR adds two features which will be useful for operators when ACLs are in use. 1. Tokens set in configuration files are now reloadable. 2. If `acl.enable_token_persistence` is set to `true` in the configuration, tokens set via the `v1/agent/token` endpoint are now persisted to disk and loaded when the agent starts (or during configuration reload) Note that token persistence is opt-in so our users who do not want tokens on the local disk will see no change. Some other secondary changes: * Refactored a bunch of places where the replication token is retrieved from the token store. This token isn't just for replicating ACLs and now it is named accordingly. * Allowed better paths in the `v1/agent/token/` API. Instead of paths like: `v1/agent/token/acl_replication_token` the path can now be just `v1/agent/token/replication`. The old paths remain to be valid. * Added a couple new API functions to set tokens via the new paths. Deprecated the old ones and pointed to the new names. The names are also generally better and don't imply that what you are setting is for ACLs but rather are setting ACL tokens. There is a minor semantic difference there especially for the replication token as again, its no longer used only for ACL token/policy replication. The new functions will detect 404s and fallback to using the older token paths when talking to pre-1.4.3 agents. * Docs updated to reflect the API additions and to show using the new endpoints. * Updated the ACL CLI set-agent-tokens command to use the non-deprecated APIs.
2019-02-27 19:28:31 +00:00
// ACLReplicationToken is used to replicate data locally from the
// PrimaryDatacenter. Replication is only available on servers in
// datacenters other than the PrimaryDatacenter
//
// DEPRECATED (ACL-Legacy-Compat): Setting this to a non-empty value
// also enables legacy ACL replication if ACLs are enabled and in legacy mode.
2017-10-18 15:23:56 +00:00
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.tokens.replication = string
2017-10-18 15:23:56 +00:00
ACLReplicationToken string
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// ACLtokenReplication is used to indicate that both tokens and policies
// should be replicated instead of just policies
//
// hcl: acl.token_replication = boolean
ACLTokenReplication bool
// ACLTokenTTL is used to control the time-to-live of cached ACL tokens. This has
2017-10-18 15:23:56 +00:00
// a major impact on performance. By default, it is set to 30 seconds.
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.policy_ttl = "duration"
ACLTokenTTL time.Duration
// ACLPolicyTTL is used to control the time-to-live of cached ACL policies. This has
// a major impact on performance. By default, it is set to 30 seconds.
//
// hcl: acl.token_ttl = "duration"
ACLPolicyTTL time.Duration
2017-10-18 15:23:56 +00:00
// ACLRoleTTL is used to control the time-to-live of cached ACL roles. This has
// a major impact on performance. By default, it is set to 30 seconds.
//
// hcl: acl.role_ttl = "duration"
ACLRoleTTL time.Duration
2017-10-18 15:23:56 +00:00
// ACLToken is the default token used to make requests if a per-request
// token is not provided. If not configured the 'anonymous' token is used.
//
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// hcl: acl.tokens.default = string
2017-10-18 15:23:56 +00:00
ACLToken string
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
ACL Token Persistence and Reloading (#5328) This PR adds two features which will be useful for operators when ACLs are in use. 1. Tokens set in configuration files are now reloadable. 2. If `acl.enable_token_persistence` is set to `true` in the configuration, tokens set via the `v1/agent/token` endpoint are now persisted to disk and loaded when the agent starts (or during configuration reload) Note that token persistence is opt-in so our users who do not want tokens on the local disk will see no change. Some other secondary changes: * Refactored a bunch of places where the replication token is retrieved from the token store. This token isn't just for replicating ACLs and now it is named accordingly. * Allowed better paths in the `v1/agent/token/` API. Instead of paths like: `v1/agent/token/acl_replication_token` the path can now be just `v1/agent/token/replication`. The old paths remain to be valid. * Added a couple new API functions to set tokens via the new paths. Deprecated the old ones and pointed to the new names. The names are also generally better and don't imply that what you are setting is for ACLs but rather are setting ACL tokens. There is a minor semantic difference there especially for the replication token as again, its no longer used only for ACL token/policy replication. The new functions will detect 404s and fallback to using the older token paths when talking to pre-1.4.3 agents. * Docs updated to reflect the API additions and to show using the new endpoints. * Updated the ACL CLI set-agent-tokens command to use the non-deprecated APIs.
2019-02-27 19:28:31 +00:00
// ACLEnableTokenPersistence determines whether or not tokens set via the agent HTTP API
// should be persisted to disk and reloaded when an agent restarts.
ACLEnableTokenPersistence bool
2017-10-18 10:32:39 +00:00
// AutopilotCleanupDeadServers enables the automatic cleanup of dead servers when new ones
// are added to the peer list. Defaults to true.
//
// hcl: autopilot { cleanup_dead_servers = (true|false) }
AutopilotCleanupDeadServers bool
// AutopilotDisableUpgradeMigration will disable Autopilot's upgrade migration
// strategy of waiting until enough newer-versioned servers have been added to the
// cluster before promoting them to voters. (Enterprise-only)
//
// hcl: autopilot { disable_upgrade_migration = (true|false)
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
AutopilotDisableUpgradeMigration bool
2017-10-18 10:32:39 +00:00
// AutopilotLastContactThreshold is the limit on the amount of time a server can go
// without leader contact before being considered unhealthy.
//
// hcl: autopilot { last_contact_threshold = "duration" }
AutopilotLastContactThreshold time.Duration
// AutopilotMaxTrailingLogs is the amount of entries in the Raft Log that a server can
// be behind before being considered unhealthy. The value must be positive.
//
// hcl: autopilot { max_trailing_logs = int }
AutopilotMaxTrailingLogs int
// AutopilotMinQuorum sets the minimum number of servers required in a cluster
// before autopilot can prune dead servers.
//
//hcl: autopilot { min_quorum = int }
AutopilotMinQuorum uint
2017-10-18 10:32:39 +00:00
// AutopilotRedundancyZoneTag is the Meta tag to use for separating servers
// into zones for redundancy. If left blank, this feature will be disabled.
// (Enterprise-only)
//
// hcl: autopilot { redundancy_zone_tag = string }
AutopilotRedundancyZoneTag string
// AutopilotServerStabilizationTime is the minimum amount of time a server must be
// in a stable, healthy state before it can be added to the cluster. Only
// applicable with Raft protocol version 3 or higher.
//
// hcl: autopilot { server_stabilization_time = "duration" }
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
AutopilotServerStabilizationTime time.Duration
2017-10-18 10:32:39 +00:00
// AutopilotUpgradeVersionTag is the node tag to use for version info when
// performing upgrade migrations. If left blank, the Consul version will be used.
//
// (Enterprise-only)
2017-10-18 10:32:39 +00:00
//
// hcl: autopilot { upgrade_version_tag = string }
AutopilotUpgradeVersionTag string
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
2017-10-18 10:20:26 +00:00
// DNSAllowStale is used to enable lookups with stale
// data. This gives horizontal read scalability since
// any Consul server can service the query instead of
// only the leader.
//
// hcl: dns_config { allow_stale = (true|false) }
DNSAllowStale bool
// DNSARecordLimit is used to limit the maximum number of DNS Resource
// Records returned in the ANSWER section of a DNS response for A or AAAA
// records for both UDP and TCP queries.
//
// This is not normally useful and will be limited based on the querying
// protocol, however systems that implemented §6 Rule 9 in RFC3484
// may want to set this to `1` in order to subvert §6 Rule 9 and
// re-obtain the effect of randomized resource records (i.e. each
// answer contains only one IP, but the IP changes every request).
// RFC3484 sorts answers in a deterministic order, which defeats the
// purpose of randomized DNS responses. This RFC has been obsoleted
// by RFC6724 and restores the desired behavior of randomized
// responses, however a large number of Linux hosts using glibc(3)
// implemented §6 Rule 9 and may need this option (e.g. CentOS 5-6,
// Debian Squeeze, etc).
//
// hcl: dns_config { a_record_limit = int }
DNSARecordLimit int
2017-10-18 10:20:26 +00:00
// DNSDisableCompression is used to control whether DNS responses are
// compressed. In Consul 0.7 this was turned on by default and this
// config was added as an opt-out.
//
// hcl: dns_config { disable_compression = (true|false) }
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
DNSDisableCompression bool
2017-10-18 10:20:26 +00:00
// DNSDomain is the DNS domain for the records. Should end with a dot.
// Defaults to "consul."
//
// hcl: domain = string
// flag: -domain string
DNSDomain string
// DNSAltDomain can be set to support resolution on an additional
// consul domain. Should end with a dot.
// If left blank, only the primary domain will be used.
//
// hcl: alt_domain = string
// flag: -alt-domain string
DNSAltDomain string
2017-10-18 10:20:26 +00:00
// DNSEnableTruncate is used to enable setting the truncate
// flag for UDP DNS queries. This allows unmodified
// clients to re-query the consul server using TCP
// when the total number of records exceeds the number
// returned by default for UDP.
//
// hcl: dns_config { enable_truncate = (true|false) }
DNSEnableTruncate bool
// DNSMaxStale is used to bound how stale of a result is
// accepted for a DNS lookup. This can be used with
// AllowStale to limit how old of a value is served up.
// If the stale result exceeds this, another non-stale
// stale read is performed.
//
// hcl: dns_config { max_stale = "duration" }
DNSMaxStale time.Duration
// DNSNodeTTL provides the TTL value for a node query.
//
// hcl: dns_config { node_ttl = "duration" }
DNSNodeTTL time.Duration
// DNSOnlyPassing is used to determine whether to filter nodes
// whose health checks are in any non-passing state. By
// default, only nodes in a critical state are excluded.
//
// hcl: dns_config { only_passing = "duration" }
DNSOnlyPassing bool
// DNSRecursorTimeout specifies the timeout in seconds
// for Consul's internal dns client used for recursion.
// This value is used for the connection, read and write timeout.
//
// hcl: dns_config { recursor_timeout = "duration" }
DNSRecursorTimeout time.Duration
// DNSServiceTTL provides the TTL value for a service
// query for given service. The "*" wildcard can be used
// to set a default for all services.
//
// hcl: dns_config { service_ttl = map[string]"duration" }
DNSServiceTTL map[string]time.Duration
// DNSUDPAnswerLimit is used to limit the maximum number of DNS Resource
// Records returned in the ANSWER section of a DNS response for UDP
// responses without EDNS support (limited to 512 bytes).
// This parameter is deprecated, if you want to limit the number of
// records returned by A or AAAA questions, please use DNSARecordLimit
// instead.
2017-10-18 10:20:26 +00:00
//
// hcl: dns_config { udp_answer_limit = int }
DNSUDPAnswerLimit int
// DNSNodeMetaTXT controls whether DNS queries will synthesize
// TXT records for the node metadata and add them when not specifically
// request (query type = TXT). If unset this will default to true
DNSNodeMetaTXT bool
2017-10-18 10:20:26 +00:00
// DNSRecursors can be set to allow the DNS servers to recursively
// resolve non-consul domains.
//
// hcl: recursors = []string
// flag: -recursor string [-recursor string]
DNSRecursors []string
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
2020-01-27 13:00:33 +00:00
// DNSUseCache whether or not to use cache for dns queries
//
// hcl: dns_config { use_cache = (true|false) }
DNSUseCache bool
2020-01-27 13:00:33 +00:00
// DNSUseCache whether or not to use cache for dns queries
//
// hcl: dns_config { cache_max_age = "duration" }
DNSCacheMaxAge time.Duration
2017-10-18 10:20:10 +00:00
// HTTPBlockEndpoints is a list of endpoint prefixes to block in the
// HTTP API. Any requests to these will get a 403 response.
//
// hcl: http_config { block_endpoints = []string }
HTTPBlockEndpoints []string
// AllowWriteHTTPFrom restricts the agent write endpoints to the given
// networks. Any request to a protected endpoint that is not mactched
// by one of these networks will get a 403 response.
// An empty slice means no restriction.
//
// hcl: http_config { allow_write_http_from = []string }
AllowWriteHTTPFrom []*net.IPNet
2017-10-18 10:20:10 +00:00
// HTTPResponseHeaders are used to add HTTP header response fields to the HTTP API responses.
//
// hcl: http_config { response_headers = map[string]string }
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
HTTPResponseHeaders map[string]string
2018-06-14 12:52:48 +00:00
// Embed Telemetry Config
Telemetry lib.TelemetryConfig
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
2017-10-19 07:47:22 +00:00
// Datacenter is the datacenter this node is in. Defaults to "dc1".
//
// Datacenter is exposed via /v1/agent/self from here and
// used in lots of places like CLI commands. Treat this as an interface
// that must be stable.
2017-10-19 07:47:22 +00:00
//
// hcl: datacenter = string
// flag: -datacenter string
Datacenter string
2017-10-19 07:47:22 +00:00
// Defines the maximum stale value for discovery path. Defaults to "0s".
// Discovery paths are /v1/heath/ paths
//
// If not set to 0, it will try to perform stale read and perform only a
// consistent read whenever the value is too old.
// hcl: discovery_max_stale = "duration"
DiscoveryMaxStale time.Duration
2017-10-19 07:47:22 +00:00
// Node name is the name we use to advertise. Defaults to hostname.
//
// NodeName is exposed via /v1/agent/self from here and
// used in lots of places like CLI commands. Treat this as an interface
// that must be stable.
//
// hcl: node_name = string
// flag: -node string
NodeName string
// AdvertiseAddrLAN is the address we use for advertising our Serf, and
// Consul RPC IP. The address can be specified as an ip address or as a
// go-sockaddr template which resolves to a single ip address. If not
// specified, the bind address is used.
//
// hcl: advertise_addr = string
2017-10-18 15:31:00 +00:00
AdvertiseAddrLAN *net.IPAddr
// AdvertiseAddrWAN is the address we use for advertising our Serf, and
// Consul RPC IP. The address can be specified as an ip address or as a
// go-sockaddr template which resolves to a single ip address. If not
// specified, the bind address is used.
//
// hcl: advertise_addr_wan = string
2017-10-18 15:31:00 +00:00
AdvertiseAddrWAN *net.IPAddr
2017-10-19 07:47:22 +00:00
// BindAddr is used to control the address we bind to.
// If not specified, the first private IP we find is used.
// This controls the address we use for cluster facing
// services (Gossip, Server RPC)
//
// The value can be either an ip address or a go-sockaddr
// template which resolves to a single ip address.
//
// hcl: bind_addr = string
// flag: -bind string
BindAddr *net.IPAddr
2017-10-18 15:31:00 +00:00
// Bootstrap is used to bring up the first Consul server, and
// permits that node to elect itself leader
//
// hcl: bootstrap = (true|false)
// flag: -bootstrap
Bootstrap bool
2017-10-23 06:05:47 +00:00
// BootstrapExpect tries to automatically bootstrap the Consul cluster, by
// having servers wait to bootstrap until enough servers join, and then
// performing the bootstrap process automatically. They will disable their
// automatic bootstrap process if they detect any servers that are part of
// an existing cluster, so it's safe to leave this set to a non-zero value.
2017-10-18 15:31:00 +00:00
//
// hcl: bootstrap_expect = int
// flag: -bootstrap-expect=int
BootstrapExpect int
2017-10-19 07:47:22 +00:00
// CAFile is a path to a certificate authority file. This is used with
// VerifyIncoming or VerifyOutgoing to verify the TLS connection.
//
// hcl: ca_file = string
CAFile string
// CAPath is a path to a directory of certificate authority files. This is
// used with VerifyIncoming or VerifyOutgoing to verify the TLS connection.
//
// hcl: ca_path = string
CAPath string
// CertFile is used to provide a TLS certificate that is used for serving
// TLS connections. Must be provided to serve TLS connections.
//
// hcl: cert_file = string
CertFile string
2017-10-19 19:26:02 +00:00
// CheckUpdateInterval controls the interval on which the output of a health check
// is updated if there is no change to the state. For example, a check in a steady
// state may run every 5 second generating a unique output (timestamp, etc), forcing
// constant writes. This allows Consul to defer the write for some period of time,
// reducing the write pressure when the state is steady.
//
// See also: DiscardCheckOutput
//
// hcl: check_update_interval = "duration"
2017-10-18 15:31:00 +00:00
CheckUpdateInterval time.Duration
2017-10-19 19:26:02 +00:00
// Maximum size for the output of a healtcheck
// hcl check_output_max_size int
// flag: -check_output_max_size int
CheckOutputMaxSize int
// Checks contains the provided check definitions.
//
// hcl: checks = [
// {
// id = string
// name = string
// notes = string
// service_id = string
// token = string
// status = string
// script = string
// args = string
// http = string
// header = map[string][]string
// method = string
// tcp = string
// interval = string
// docker_container_id = string
// shell = string
// tls_skip_verify = (true|false)
// timeout = "duration"
// ttl = "duration"
// deregister_critical_service_after = "duration"
// },
// ...
// ]
2017-10-19 19:26:02 +00:00
Checks []*structs.CheckDefinition
// ClientAddrs contains the list of ip addresses the DNS, HTTP and HTTPS
// endpoints will bind to if the endpoints are enabled (ports > 0) and the
// addresses are not overwritten.
//
// The ip addresses must be provided as a space separated list of ip
// addresses and go-sockaddr templates.
//
// Client addresses cannot contain UNIX socket addresses since a socket
// cannot be shared across multiple endpoints (no ports). To use UNIX
// sockets configure it in 'addresses'.
//
// hcl: client_addr = string
// flag: -client string
ClientAddrs []*net.IPAddr
// ConfigEntryBootstrap contains a list of ConfigEntries to ensure are created
// If entries of the same Kind/Name exist already these will not update them.
ConfigEntryBootstrap []structs.ConfigEntry
// AutoEncryptTLS requires the client to acquire TLS certificates from
// servers.
AutoEncryptTLS bool
// Additional DNS SAN entries that clients request during auto_encrypt
// flow for their certificates.
AutoEncryptDNSSAN []string
// Additional IP SAN entries that clients request during auto_encrypt
// flow for their certificates.
AutoEncryptIPSAN []net.IP
// AutoEncryptAllowTLS enables the server to respond to
// AutoEncrypt.Sign requests.
AutoEncryptAllowTLS bool
// ConnectEnabled opts the agent into connect. It should be set on all clients
2018-06-07 09:17:44 +00:00
// and servers in a cluster for correct connect operation.
ConnectEnabled bool
// ConnectSidecarMinPort is the inclusive start of the range of ports
// allocated to the agent for asigning to sidecar services where no port is
// specified.
ConnectSidecarMinPort int
// ConnectSidecarMaxPort is the inclusive end of the range of ports
// allocated to the agent for asigning to sidecar services where no port is
// specified
ConnectSidecarMaxPort int
// ExposeMinPort is the inclusive start of the range of ports
// allocated to the agent for exposing checks through a proxy
ExposeMinPort int
// ExposeMinPort is the inclusive start of the range of ports
// allocated to the agent for exposing checks through a proxy
ExposeMaxPort int
// ConnectCAProvider is the type of CA provider to use with Connect.
ConnectCAProvider string
// ConnectCAConfig is the config to use for the CA provider.
ConnectCAConfig map[string]interface{}
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
// ConnectMeshGatewayWANFederationEnabled determines if wan federation of
// datacenters should exclusively traverse mesh gateways.
ConnectMeshGatewayWANFederationEnabled bool
// ConnectTestCALeafRootChangeSpread is used to control how long the CA leaf
// cache with spread CSRs over when a root change occurs. For now we don't
// expose this in public config intentionally but could later with a rename.
// We only set this from during tests to effectively make CA rotation tests
// deterministic again.
ConnectTestCALeafRootChangeSpread time.Duration
2017-10-19 19:26:02 +00:00
// DNSAddrs contains the list of TCP and UDP addresses the DNS server will
// bind to. If the DNS endpoint is disabled (ports.dns <= 0) the list is
// empty.
//
// The ip addresses are taken from 'addresses.dns' which should contain a
// space separated list of ip addresses and/or go-sockaddr templates.
//
// If 'addresses.dns' was not provided the 'client_addr' addresses are
// used.
//
// The DNS server cannot be bound to UNIX sockets.
//
// hcl: client_addr = string addresses { dns = string } ports { dns = int }
DNSAddrs []net.Addr
// DNSPort is the port the DNS server listens on. The default is 8600.
// Setting this to a value <= 0 disables the endpoint.
//
// hcl: ports { dns = int }
// flags: -dns-port int
DNSPort int
2017-10-19 07:47:22 +00:00
// DNSSOA is the settings applied for DNS SOA
// hcl: soa {}
DNSSOA RuntimeSOAConfig
2017-10-19 07:47:22 +00:00
// DataDir is the path to the directory where the local state is stored.
//
// hcl: data_dir = string
// flag: -data-dir string
DataDir string
2017-10-18 15:31:00 +00:00
// DefaultQueryTime is the amount of time a blocking query will wait before
// Consul will force a response. This value can be overridden by the 'wait'
// query parameter.
//
// hcl: default_query_time = "duration"
// flag: -default-query-time string
DefaultQueryTime time.Duration
2017-10-18 15:31:00 +00:00
// DevMode enables a fast-path mode of operation to bring up an in-memory
// server with minimal configuration. Useful for developing Consul.
//
// flag: -dev
DevMode bool
2017-10-19 19:26:02 +00:00
// DisableAnonymousSignature is used to turn off the anonymous signature
// send with the update check. This is used to deduplicate messages.
//
// hcl: disable_anonymous_signature = (true|false)
2017-10-18 15:23:56 +00:00
DisableAnonymousSignature bool
2017-10-19 19:26:02 +00:00
// DisableCoordinates controls features related to network coordinates.
//
// hcl: disable_coordinates = (true|false)
DisableCoordinates bool
2017-10-19 07:47:22 +00:00
// DisableHostNodeID will prevent Consul from using information from the
// host to generate a node ID, and will cause Consul to generate a
// random ID instead.
//
// hcl: disable_host_node_id = (true|false)
2017-10-19 19:26:02 +00:00
// flag: -disable-host-node-id
2017-10-19 07:47:22 +00:00
DisableHostNodeID bool
// DisableHTTPUnprintableCharFilter will bypass the filter preventing HTTP
// URLs from containing unprintable chars. This filter was added in 1.0.3 as a
// response to a vulnerability report. Disabling this is never recommended in
// general however some users who have keys written in older versions of
// Consul may use this to temporarily disable the filter such that they can
// delete those keys again! We do not recommend leaving it disabled long term.
//
// hcl: disable_http_unprintable_char_filter
DisableHTTPUnprintableCharFilter bool
2017-10-19 07:47:22 +00:00
// DisableKeyringFile disables writing the keyring to a file.
//
// hcl: disable_keyring_file = (true|false)
// flag: -disable-keyring-file
DisableKeyringFile bool
2017-10-19 19:26:02 +00:00
// DisableRemoteExec is used to turn off the remote execution
// feature. This is for security to prevent unknown scripts from running.
//
// hcl: disable_remote_exec = (true|false)
DisableRemoteExec bool
// DisableUpdateCheck is used to turn off the automatic update and
// security bulletin checking.
//
// hcl: disable_update_check = (true|false)
2017-10-19 07:47:22 +00:00
DisableUpdateCheck bool
2017-10-19 19:26:02 +00:00
// DiscardCheckOutput is used to turn off storing and comparing the
// output of health checks. This reduces the write rate on the server
// for checks with highly volatile output. (reloadable)
//
// See also: CheckUpdateInterval
//
// hcl: discard_check_output = (true|false)
2017-10-19 07:47:22 +00:00
DiscardCheckOutput bool
2017-10-18 15:23:56 +00:00
// EnableAgentTLSForChecks is used to apply the agent's TLS settings in
// order to configure the HTTP client used for health checks. Enabling
// this allows HTTP checks to present a client certificate and verify
// the server using the same TLS configuration as the agent (CA, cert,
// and key).
EnableAgentTLSForChecks bool
// EnableCentralServiceConfig controls whether the agent should incorporate
// centralized config such as service-defaults into local service registrations.
//
2019-04-25 09:11:07 +00:00
// hcl: enable_central_service_config = (true|false)
EnableCentralServiceConfig bool
2017-10-19 07:47:22 +00:00
// EnableDebug is used to enable various debugging features.
//
// hcl: enable_debug = (true|false)
EnableDebug bool
// EnableLocalScriptChecks controls whether health checks declared from the local
// config file which execute scripts are enabled. This includes regular script
// checks and Docker checks.
//
// hcl: (enable_script_checks|enable_local_script_checks) = (true|false)
// flag: -enable-script-checks, -enable-local-script-checks
EnableLocalScriptChecks bool
// EnableRemoeScriptChecks controls whether health checks declared from the http API
// which execute scripts are enabled. This includes regular script checks and Docker
2017-10-19 19:26:02 +00:00
// checks.
//
// hcl: enable_script_checks = (true|false)
// flag: -enable-script-checks
EnableRemoteScriptChecks bool
2017-10-19 19:26:02 +00:00
// EnableSyslog is used to also tee all the logs over to syslog. Only supported
// on linux and OSX. Other platforms will generate an error.
//
// hcl: enable_syslog = (true|false)
// flag: -syslog
EnableSyslog bool
// EnableUI enables the statically-compiled assets for the Consul web UI and
// serves them at the default /ui/ endpoint automatically.
//
// hcl: enable_ui = (true|false)
// flag: -ui
EnableUI bool
2017-10-19 07:47:22 +00:00
// EncryptKey contains the encryption key to use for the Serf communication.
//
// hcl: encrypt = string
// flag: -encrypt string
EncryptKey string
// EncryptVerifyIncoming enforces incoming gossip encryption and can be
// used to upshift to encrypted gossip on a running cluster.
//
// hcl: encrypt_verify_incoming = (true|false)
2017-10-18 15:31:00 +00:00
EncryptVerifyIncoming bool
2017-10-19 07:47:22 +00:00
// EncryptVerifyOutgoing enforces outgoing gossip encryption and can be
// used to upshift to encrypted gossip on a running cluster.
//
// hcl: encrypt_verify_outgoing = (true|false)
2017-10-18 15:31:00 +00:00
EncryptVerifyOutgoing bool
2017-10-19 07:47:22 +00:00
// GRPCPort is the port the gRPC server listens on. Currently this only
// exposes the xDS and ext_authz APIs for Envoy and it is disabled by default.
//
// hcl: ports { grpc = int }
// flags: -grpc-port int
GRPCPort int
// GRPCAddrs contains the list of TCP addresses and UNIX sockets the gRPC
// server will bind to. If the gRPC endpoint is disabled (ports.grpc <= 0)
// the list is empty.
//
// The addresses are taken from 'addresses.grpc' which should contain a
// space separated list of ip addresses, UNIX socket paths and/or
// go-sockaddr templates. UNIX socket paths must be written as
// 'unix://<full path>', e.g. 'unix:///var/run/consul-grpc.sock'.
//
// If 'addresses.grpc' was not provided the 'client_addr' addresses are
// used.
//
// hcl: client_addr = string addresses { grpc = string } ports { grpc = int }
GRPCAddrs []net.Addr
// HTTPAddrs contains the list of TCP addresses and UNIX sockets the HTTP
// server will bind to. If the HTTP endpoint is disabled (ports.http <= 0)
// the list is empty.
//
// The addresses are taken from 'addresses.http' which should contain a
// space separated list of ip addresses, UNIX socket paths and/or
// go-sockaddr templates. UNIX socket paths must be written as
// 'unix://<full path>', e.g. 'unix:///var/run/consul-http.sock'.
//
// If 'addresses.http' was not provided the 'client_addr' addresses are
// used.
//
// hcl: client_addr = string addresses { http = string } ports { http = int }
2017-10-19 19:26:02 +00:00
HTTPAddrs []net.Addr
// HTTPPort is the port the HTTP server listens on. The default is 8500.
// Setting this to a value <= 0 disables the endpoint.
//
// hcl: ports { http = int }
// flags: -http-port int
HTTPPort int
// HTTPSAddrs contains the list of TCP addresses and UNIX sockets the HTTPS
// server will bind to. If the HTTPS endpoint is disabled (ports.https <=
// 0) the list is empty.
//
// The addresses are taken from 'addresses.https' which should contain a
// space separated list of ip addresses, UNIX socket paths and/or
// go-sockaddr templates. UNIX socket paths must be written as
// 'unix://<full path>', e.g. 'unix:///var/run/consul-https.sock'.
//
// If 'addresses.https' was not provided the 'client_addr' addresses are
// used.
//
// hcl: client_addr = string addresses { https = string } ports { https = int }
2017-10-19 07:47:22 +00:00
HTTPSAddrs []net.Addr
2017-10-19 19:26:02 +00:00
// HTTPMaxConnsPerClient limits the number of concurrent TCP connections the
// HTTP(S) server will accept from any single source IP address.
//
// hcl: limits{ http_max_conns_per_client = 200 }
HTTPMaxConnsPerClient int
// HTTPSHandshakeTimeout is the time allowed for HTTPS client to complete the
// TLS handshake and send first bytes of the request.
//
// hcl: limits{ https_handshake_timeout = "5s" }
HTTPSHandshakeTimeout time.Duration
2017-10-19 19:26:02 +00:00
// HTTPSPort is the port the HTTP server listens on. The default is -1.
// Setting this to a value <= 0 disables the endpoint.
//
// hcl: ports { https = int }
// flags: -https-port int
2017-10-19 19:26:02 +00:00
HTTPSPort int
2017-10-19 07:47:22 +00:00
// KeyFile is used to provide a TLS key that is used for serving TLS
// connections. Must be provided to serve TLS connections.
//
// hcl: key_file = string
KeyFile string
// KVMaxValueSize controls the max allowed value size. If not set defaults
// to raft's suggested max value size.
//
// hcl: limits { kv_max_value_size = uint64 }
KVMaxValueSize uint64
// LeaveDrainTime is used to wait after a server has left the LAN Serf
// pool for RPCs to drain and new requests to be sent to other servers.
2017-10-19 19:26:02 +00:00
//
// hcl: performance { leave_drain_time = "duration" }
2017-10-19 07:47:22 +00:00
LeaveDrainTime time.Duration
// LeaveOnTerm controls if Serf does a graceful leave when receiving
// the TERM signal. Defaults true on clients, false on servers. (reloadable)
//
// hcl: leave_on_terminate = (true|false)
LeaveOnTerm bool
// LogLevel is the level of the logs to write. Defaults to "INFO".
//
// hcl: log_level = string
LogLevel string
// LogJSON controls whether to output logs as structured JSON. Defaults to false.
//
// hcl: log_json = (true|false)
// flag: -log-json
LogJSON bool
// LogFile is the path to the file where the logs get written to. Defaults to empty string.
//
// hcl: log_file = string
// flags: -log-file string
LogFile string
// LogRotateDuration is the time configured to rotate logs based on time
//
// hcl: log_rotate_duration = string
// flags: -log-rotate-duration string
LogRotateDuration time.Duration
// LogRotateBytes is the time configured to rotate logs based on bytes written
//
// hcl: log_rotate_bytes = int
// flags: -log-rotate-bytes int
LogRotateBytes int
// LogRotateMaxFiles is the maximum number of log file archives to keep
//
// hcl: log_rotate_max_files = int
// flags: -log-rotate-max-files int
LogRotateMaxFiles int
// MaxQueryTime is the maximum amount of time a blocking query can wait
// before Consul will force a response. Consul applies jitter to the wait
// time. The jittered time will be capped to MaxQueryTime.
//
// hcl: max_query_time = "duration"
// flags: -max-query-time string
MaxQueryTime time.Duration
2017-10-19 07:47:22 +00:00
// Node ID is a unique ID for this node across space and time. Defaults
// to a randomly-generated ID that persists in the data-dir.
//
// todo(fs): don't we have a requirement for this to be a UUID in a specific format?
//
// hcl: node_id = string
// flag: -node-id string
NodeID types.NodeID
// NodeMeta contains metadata key/value pairs. These are excluded from JSON output
// because they can be reloaded and might be stale when shown from the
// config instead of the local state.
// todo(fs): should the sanitizer omit them from output as well since they could be stale?
//
// hcl: node_meta = map[string]string
// flag: -node-meta "key:value" -node-meta "key:value" ...
NodeMeta map[string]string
// NonVotingServer is whether this server will act as a non-voting member
// of the cluster to help provide read scalability. (Enterprise-only)
//
// hcl: non_voting_server = (true|false)
// flag: -non-voting-server
NonVotingServer bool
2017-10-19 19:26:02 +00:00
// PidFile is the file to store our PID in.
//
// hcl: pid_file = string
PidFile string
// PrimaryDatacenter is the central datacenter that holds authoritative
// ACL records, replicates intentions and holds the root CA for Connect.
// This must be the same for the entire cluster. Off by default.
//
// hcl: primary_datacenter = string
PrimaryDatacenter string
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
// PrimaryGateways is a list of addresses and/or go-discover expressions to
// discovery the mesh gateways in the primary datacenter. See
// https://www.consul.io/docs/agent/options.html#cloud-auto-joining for
// details.
//
// hcl: primary_gateways = []string
// flag: -primary-gateway string -primary-gateway string
PrimaryGateways []string
// PrimaryGatewaysInterval specifies the amount of time to wait in between discovery
// attempts on agent start. The minimum allowed value is 1 second and
// the default is 30s.
//
// hcl: primary_gateways_interval = "duration"
PrimaryGatewaysInterval time.Duration
// RPCAdvertiseAddr is the TCP address Consul advertises for its RPC endpoint.
// By default this is the bind address on the default RPC Server port. If the
// advertise address is specified then it is used.
//
// hcl: bind_addr = string advertise_addr = string ports { server = int }
2017-10-19 07:47:22 +00:00
RPCAdvertiseAddr *net.TCPAddr
2017-10-19 19:26:02 +00:00
// RPCBindAddr is the TCP address Consul will bind to for its RPC endpoint.
// By default this is the bind address on the default RPC Server port.
2017-10-19 19:26:02 +00:00
//
// hcl: bind_addr = string ports { server = int }
RPCBindAddr *net.TCPAddr
// RPCHandshakeTimeout is the timeout for reading the initial magic byte on a
// new RPC connection. If this is set high it may allow unauthenticated users
// to hold connections open arbitrarily long, even when mutual TLS is being
// enforced. It may be set to 0 explicitly to disable the timeout but this
// should never be used in production. Default is 5 seconds.
//
// hcl: limits { rpc_handshake_timeout = "duration" }
RPCHandshakeTimeout time.Duration
// RPCHoldTimeout is how long an RPC can be "held" before it is errored.
// This is used to paper over a loss of leadership by instead holding RPCs,
// so that the caller experiences a slow response rather than an error.
// This period is meant to be long enough for a leader election to take
// place, and a small jitter is applied to avoid a thundering herd.
//
// hcl: performance { rpc_hold_timeout = "duration" }
2017-10-19 19:26:02 +00:00
RPCHoldTimeout time.Duration
2017-10-18 15:31:00 +00:00
// RPCRateLimit and RPCMaxBurst control how frequently RPC calls are allowed
// to happen. In any large enough time interval, rate limiter limits the
// rate to RPCRate tokens per second, with a maximum burst size of
// RPCMaxBurst events. As a special case, if RPCRate == Inf (the infinite
// rate), RPCMaxBurst is ignored.
//
// See https://en.wikipedia.org/wiki/Token_bucket for more about token
// buckets.
//
// hcl: limit { rpc_rate = (float64|MaxFloat64) rpc_max_burst = int }
RPCRateLimit rate.Limit
RPCMaxBurst int
// RPCMaxConnsPerClient limits the number of concurrent TCP connections the
// RPC server will accept from any single source IP address.
//
// hcl: limits{ rpc_max_conns_per_client = 100 }
RPCMaxConnsPerClient int
2017-10-19 07:47:22 +00:00
// RPCProtocol is the Consul protocol version to use.
//
// hcl: protocol = int
RPCProtocol int
// RaftProtocol sets the Raft protocol version to use on this server.
// Defaults to 3.
//
// hcl: raft_protocol = int
RaftProtocol int
// RaftSnapshotThreshold sets the minimum threshold of raft commits after which
// a snapshot is created. Defaults to 8192
//
// hcl: raft_snapshot_threshold = int
RaftSnapshotThreshold int
// RaftSnapshotInterval sets the interval to use when checking whether to create
// a new snapshot. Defaults to 5 seconds.
// hcl: raft_snapshot_threshold = int
RaftSnapshotInterval time.Duration
// RaftTrailingLogs sets the number of log entries that will be left in the
// log store after a snapshot. This must be large enough that a follower can
// transfer and restore an entire snapshot of the state before this many new
// entries have been appended. In vast majority of cases the default is plenty
// but if there is a sustained high write throughput coupled with a huge
// multi-gigabyte snapshot setting this higher may be necessary to allow
// followers time to reload from snapshot without becoming unhealthy. If it's
// too low then followers are unable to ever recover from a restart and will
// enter a loop of constantly downloading full snapshots and never catching
// up. If you need to change this you should reconsider your usage of Consul
// as it is not designed to store multiple-gigabyte data sets with high write
// throughput. Defaults to 10000.
//
// hcl: raft_trailing_logs = int
RaftTrailingLogs int
2017-10-19 19:26:02 +00:00
// ReconnectTimeoutLAN specifies the amount of time to wait to reconnect with
// another agent before deciding it's permanently gone. This can be used to
// control the time it takes to reap failed nodes from the cluster.
//
// hcl: reconnect_timeout = "duration"
ReconnectTimeoutLAN time.Duration
// ReconnectTimeoutWAN specifies the amount of time to wait to reconnect with
// another agent before deciding it's permanently gone. This can be used to
// control the time it takes to reap failed nodes from the cluster.
//
// hcl: reconnect_timeout = "duration"
ReconnectTimeoutWAN time.Duration
// RejoinAfterLeave controls our interaction with the cluster after leave.
// When set to false (default), a leave causes Consul to not rejoin
// the cluster until an explicit join is received. If this is set to
// true, we ignore the leave, and rejoin the cluster on start.
//
// hcl: rejoin_after_leave = (true|false)
// flag: -rejoin
RejoinAfterLeave bool
// RetryJoinIntervalLAN specifies the amount of time to wait in between join
// attempts on agent start. The minimum allowed value is 1 second and
// the default is 30s.
//
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
// hcl: retry_interval = "duration"
2017-10-19 19:26:02 +00:00
RetryJoinIntervalLAN time.Duration
// RetryJoinIntervalWAN specifies the amount of time to wait in between join
// attempts on agent start. The minimum allowed value is 1 second and
// the default is 30s.
//
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
// hcl: retry_interval_wan = "duration"
2017-10-19 19:26:02 +00:00
RetryJoinIntervalWAN time.Duration
// RetryJoinLAN is a list of addresses and/or go-discover expressions to
// join with retry enabled. See
// https://www.consul.io/docs/agent/options.html#cloud-auto-joining for
// details.
//
// hcl: retry_join = []string
// flag: -retry-join string -retry-join string
RetryJoinLAN []string
// RetryJoinMaxAttemptsLAN specifies the maximum number of times to retry
// joining a host on startup. This is useful for cases where we know the
// node will be online eventually.
//
// hcl: retry_max = int
// flag: -retry-max int
2017-10-19 07:47:22 +00:00
RetryJoinMaxAttemptsLAN int
2017-10-19 19:26:02 +00:00
// RetryJoinMaxAttemptsWAN specifies the maximum number of times to retry
// joining a host on startup. This is useful for cases where we know the
// node will be online eventually.
//
// hcl: retry_max_wan = int
// flag: -retry-max-wan int
2017-10-19 07:47:22 +00:00
RetryJoinMaxAttemptsWAN int
2017-10-19 19:26:02 +00:00
// RetryJoinWAN is a list of addresses and/or go-discover expressions to
// join -wan with retry enabled. See
// https://www.consul.io/docs/agent/options.html#cloud-auto-joining for
// details.
//
// hcl: retry_join_wan = []string
// flag: -retry-join-wan string -retry-join-wan string
RetryJoinWAN []string
2017-10-19 07:47:22 +00:00
// SegmentName is the network segment for this client to join.
// (Enterprise-only)
//
// hcl: segment = string
SegmentName string
// Segments is the list of network segments for this server to
// initialize.
//
// hcl: segment = [
// {
// # name is the name of the segment
// name = string
//
// # bind is the bind ip address for this segment.
// bind = string
//
// # port is the bind port for this segment.
// port = int
//
// # advertise is the advertise ip address for this segment.
// # Defaults to the bind address if not set.
// advertise = string
//
// # rpc_listener controls whether or not to bind a separate
// # RPC listener to the bind address.
// rpc_listener = (true|false)
// },
// ...
// ]
Segments []structs.NetworkSegment
// SerfAdvertiseAddrLAN is the TCP address which is used for advertising
// the LAN Gossip pool for both client and server. The address is the
// combination of AdvertiseAddrLAN and the SerfPortLAN. If the advertise
// address is not given the bind address is used.
//
// hcl: bind_addr = string advertise_addr = string ports { serf_lan = int }
2017-10-19 07:47:22 +00:00
SerfAdvertiseAddrLAN *net.TCPAddr
// SerfAdvertiseAddrWAN is the TCP address which is used for advertising
// the WAN Gossip pool on the server only. The address is the combination
// of AdvertiseAddrWAN and the SerfPortWAN. If the advertise address is not
// given the bind address is used.
//
// hcl: bind_addr = string advertise_addr_wan = string ports { serf_wan = int }
2017-10-19 07:47:22 +00:00
SerfAdvertiseAddrWAN *net.TCPAddr
// SerfBindAddrLAN is the address to bind the Serf LAN TCP and UDP
// listeners to. The ip address is either the default bind address or the
// 'serf_lan' address which can be either an ip address or a go-sockaddr
// template which resolves to a single ip address.
//
// hcl: bind_addr = string serf_lan = string ports { serf_lan = int }
// flag: -serf-lan string
SerfBindAddrLAN *net.TCPAddr
// SerfBindAddrWAN is the address to bind the Serf WAN TCP and UDP
// listeners to. The ip address is either the default bind address or the
// 'serf_wan' address which can be either an ip address or a go-sockaddr
// template which resolves to a single ip address.
//
// hcl: bind_addr = string serf_wan = string ports { serf_wan = int }
// flag: -serf-wan string
SerfBindAddrWAN *net.TCPAddr
// SerfPortLAN is the port used for the LAN Gossip pool for both client and server.
// The default is 8301.
//
// hcl: ports { serf_lan = int }
SerfPortLAN int
// SerfPortWAN is the port used for the WAN Gossip pool for the server only.
// The default is 8302.
//
// hcl: ports { serf_wan = int }
SerfPortWAN int
2017-10-19 07:47:22 +00:00
// GossipLANGossipInterval is the interval between sending messages that need
// to be gossiped that haven't been able to piggyback on probing messages.
// If this is set to zero, non-piggyback gossip is disabled. By lowering
// this value (more frequent) gossip messages are propagated across
// the cluster more quickly at the expense of increased bandwidth. This
// configuration only applies to LAN gossip communications
//
// The default is: 200ms
//
// hcl: gossip_lan { gossip_interval = duration}
GossipLANGossipInterval time.Duration
// GossipLANGossipNodes is the number of random nodes to send gossip messages to
// per GossipInterval. Increasing this number causes the gossip messages to
// propagate across the cluster more quickly at the expense of increased
// bandwidth. This configuration only applies to LAN gossip communications
//
// The default is: 3
//
// hcl: gossip_lan { gossip_nodes = int }
GossipLANGossipNodes int
// GossipLANProbeInterval is the interval between random node probes. Setting
// this lower (more frequent) will cause the memberlist cluster to detect
// failed nodes more quickly at the expense of increased bandwidth usage.
// This configuration only applies to LAN gossip communications
//
// The default is: 1s
//
// hcl: gossip_lan { probe_interval = duration }
GossipLANProbeInterval time.Duration
// GossipLANProbeTimeout is the timeout to wait for an ack from a probed node
// before assuming it is unhealthy. This should be set to 99-percentile
// of RTT (round-trip time) on your network. This configuration
// only applies to the LAN gossip communications
//
// The default is: 500ms
//
// hcl: gossip_lan { probe_timeout = duration }
GossipLANProbeTimeout time.Duration
// GossipLANSuspicionMult is the multiplier for determining the time an
// inaccessible node is considered suspect before declaring it dead. This
// configuration only applies to LAN gossip communications
//
// The actual timeout is calculated using the formula:
//
// SuspicionTimeout = SuspicionMult * log(N+1) * ProbeInterval
//
// This allows the timeout to scale properly with expected propagation
// delay with a larger cluster size. The higher the multiplier, the longer
// an inaccessible node is considered part of the cluster before declaring
// it dead, giving that suspect node more time to refute if it is indeed
// still alive.
//
// The default is: 4
//
// hcl: gossip_lan { suspicion_mult = int }
GossipLANSuspicionMult int
// GossipLANRetransmitMult is the multiplier for the number of retransmissions
// that are attempted for messages broadcasted over gossip. This
// configuration only applies to LAN gossip communications. The actual
// count of retransmissions is calculated using the formula:
//
// Retransmits = RetransmitMult * log(N+1)
//
// This allows the retransmits to scale properly with cluster size. The
// higher the multiplier, the more likely a failed broadcast is to converge
// at the expense of increased bandwidth.
//
// The default is: 4
//
// hcl: gossip_lan { retransmit_mult = int }
GossipLANRetransmitMult int
// GossipWANGossipInterval is the interval between sending messages that need
// to be gossiped that haven't been able to piggyback on probing messages.
// If this is set to zero, non-piggyback gossip is disabled. By lowering
// this value (more frequent) gossip messages are propagated across
// the cluster more quickly at the expense of increased bandwidth. This
// configuration only applies to WAN gossip communications
//
// The default is: 200ms
//
// hcl: gossip_wan { gossip_interval = duration}
GossipWANGossipInterval time.Duration
// GossipWANGossipNodes is the number of random nodes to send gossip messages to
// per GossipInterval. Increasing this number causes the gossip messages to
// propagate across the cluster more quickly at the expense of increased
// bandwidth. This configuration only applies to WAN gossip communications
//
// The default is: 3
//
// hcl: gossip_wan { gossip_nodes = int }
GossipWANGossipNodes int
// GossipWANProbeInterval is the interval between random node probes. Setting
// this lower (more frequent) will cause the memberlist cluster to detect
// failed nodes more quickly at the expense of increased bandwidth usage.
// This configuration only applies to WAN gossip communications
//
// The default is: 1s
//
// hcl: gossip_wan { probe_interval = duration }
GossipWANProbeInterval time.Duration
// GossipWANProbeTimeout is the timeout to wait for an ack from a probed node
// before assuming it is unhealthy. This should be set to 99-percentile
// of RTT (round-trip time) on your network. This configuration
// only applies to the WAN gossip communications
//
// The default is: 500ms
//
// hcl: gossip_wan { probe_timeout = duration }
GossipWANProbeTimeout time.Duration
// GossipWANSuspicionMult is the multiplier for determining the time an
// inaccessible node is considered suspect before declaring it dead. This
// configuration only applies to WAN gossip communications
//
// The actual timeout is calculated using the formula:
//
// SuspicionTimeout = SuspicionMult * log(N+1) * ProbeInterval
//
// This allows the timeout to scale properly with expected propagation
// delay with a larger cluster size. The higher the multiplier, the longer
// an inaccessible node is considered part of the cluster before declaring
// it dead, giving that suspect node more time to refute if it is indeed
// still alive.
//
// The default is: 4
//
// hcl: gossip_wan { suspicion_mult = int }
GossipWANSuspicionMult int
// GossipWANRetransmitMult is the multiplier for the number of retransmissions
// that are attempted for messages broadcasted over gossip. This
// configuration only applies to WAN gossip communications. The actual
// count of retransmissions is calculated using the formula:
//
// Retransmits = RetransmitMult * log(N+1)
//
// This allows the retransmits to scale properly with cluster size. The
// higher the multiplier, the more likely a failed broadcast is to converge
// at the expense of increased bandwidth.
//
// The default is: 4
//
// hcl: gossip_wan { retransmit_mult = int }
GossipWANRetransmitMult int
2017-10-19 07:47:22 +00:00
// ServerMode controls if this agent acts like a Consul server,
// or merely as a client. Servers have more state, take part
// in leader election, etc.
//
// hcl: server = (true|false)
// flag: -server
ServerMode bool
// ServerName is used with the TLS certificates to ensure the name we
// provide matches the certificate.
//
// hcl: server_name = string
ServerName string
// ServerPort is the port the RPC server will bind to.
// The default is 8300.
//
// hcl: ports { server = int }
2017-10-19 19:26:02 +00:00
ServerPort int
// Services contains the provided service definitions:
//
// hcl: services = [
// {
// id = string
// name = string
// tags = []string
// address = string
// check = { check definition }
// checks = [ { check definition}, ... ]
// token = string
// enable_tag_override = (true|false)
// },
// ...
// ]
Services []*structs.ServiceDefinition
2017-10-19 19:26:02 +00:00
// Minimum Session TTL.
//
// hcl: session_ttl_min = "duration"
2017-10-19 07:47:22 +00:00
SessionTTLMin time.Duration
// SkipLeaveOnInt controls if Serf skips a graceful leave when
// receiving the INT signal. Defaults false on clients, true on
// servers. (reloadable)
//
// hcl: skip_leave_on_interrupt = (true|false)
SkipLeaveOnInt bool
// StartJoinAddrsLAN is a list of addresses to attempt to join -lan when the
2017-10-19 19:26:02 +00:00
// agent starts. If Serf is unable to communicate with any of these
// addresses, then the agent will error and exit.
//
// hcl: start_join = []string
// flag: -join string -join string
2017-10-19 07:47:22 +00:00
StartJoinAddrsLAN []string
2017-10-19 19:26:02 +00:00
// StartJoinWAN is a list of addresses to attempt to join -wan when the
// agent starts. If Serf is unable to communicate with any of these
// addresses, then the agent will error and exit.
//
// hcl: start_join_wan = []string
// flag: -join-wan string -join-wan string
2017-10-19 07:47:22 +00:00
StartJoinAddrsWAN []string
2017-10-19 19:26:02 +00:00
// SyslogFacility is used to control where the syslog messages go
// By default, goes to LOCAL0
//
// hcl: syslog_facility = string
SyslogFacility string
2017-10-19 07:47:22 +00:00
// TLSCipherSuites is used to specify the list of supported ciphersuites.
//
// The values should be a list of the following values:
//
// TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA
// TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256
2017-10-19 07:47:22 +00:00
// TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
// TLS_ECDHE_ECDSA_WITH_AES_256_CBC_SHA
2017-10-19 07:47:22 +00:00
// TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384
// TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA
// TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256
// TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256
// TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA
// TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384
2017-10-19 07:47:22 +00:00
//
// todo(fs): IMHO, we should also support the raw 0xNNNN values from
// todo(fs): https://golang.org/pkg/crypto/tls/#pkg-constants
// todo(fs): since they are standardized by IANA.
//
// hcl: tls_cipher_suites = []string
TLSCipherSuites []uint16
// TLSMinVersion is used to set the minimum TLS version used for TLS
2020-02-19 22:22:31 +00:00
// connections. Should be either "tls10", "tls11", "tls12" or "tls13".
// Defaults to tls12.
2017-10-19 07:47:22 +00:00
//
// hcl: tls_min_version = string
TLSMinVersion string
// TLSPreferServerCipherSuites specifies whether to prefer the server's
// cipher suite over the client cipher suites.
//
// hcl: tls_prefer_server_cipher_suites = (true|false)
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
TLSPreferServerCipherSuites bool
2017-10-19 07:47:22 +00:00
// TaggedAddresses are used to publish a set of addresses for
// for a node, which can be used by the remote agent. We currently
// populate only the "wan" tag based on the SerfWan advertise address,
// but this structure is here for possible future features with other
// user-defined tags. The "wan" tag will be used by remote agents if
// they are configured with TranslateWANAddrs set to true.
//
// hcl: tagged_addresses = map[string]string
TaggedAddresses map[string]string
// TranslateWANAddrs controls whether or not Consul should prefer
// the "wan" tagged address when doing lookups in remote datacenters.
// See TaggedAddresses below for more details.
//
// hcl: translate_wan_addrs = (true|false)
TranslateWANAddrs bool
// TxnMaxReqLen configures the upper limit for the size (in bytes) of the
// incoming request bodies for transactions to the /txn endpoint.
//
// hcl: limits { txn_max_req_len = uint64 }
TxnMaxReqLen uint64
2017-10-19 19:26:02 +00:00
// UIDir is the directory containing the Web UI resources.
// If provided, the UI endpoints will be enabled.
//
// hcl: ui_dir = string
// flag: -ui-dir string
UIDir string
ui: modify content path (#5950) * Add ui-content-path flag * tests complete, regex validator on string, index.html updated * cleaning up debugging stuff * ui: Enable ember environment configuration to be set via the go binary at runtime (#5934) * ui: Only inject {{.ContentPath}} if we are makeing a prod build... ...otherwise we just use the current rootURL This gets injected into a <base /> node which solves the assets path problem but not the ember problem * ui: Pull out the <base href=""> value and inject it into ember env See previous commit: The <base href=""> value is 'sometimes' injected from go at index serve time. We pass this value down to ember by overwriting the ember config that is injected via a <meta> tag. This has to be done before ember bootup. Sometimes (during testing and development, basically not production) this is injected with the already existing value, in which case this essentially changes nothing. The code here is slightly abstracted away from our specific usage to make it easier for anyone else to use, and also make sure we can cope with using this same method to pass variables down from the CLI through to ember in the future. * ui: We can't use <base /> move everything to javascript (#5941) Unfortuantely we can't seem to be able to use <base> and rootURL together as URL paths will get doubled up (`ui/ui/`). This moves all the things that we need to interpolate with .ContentPath to the `startup` javascript so we can conditionally print out `{{.ContentPath}}` in lots of places (now we can't use base) * fixed when we serve index.html * ui: For writing a ContentPath, we also need to cope with testing... (#5945) ...and potentially more environments Testing has more additional things in a separate index.html in `tests/` This make the entire thing a little saner and uses just javascriopt template literals instead of a pseudo handbrake synatx for our templating of these files. Intead of just templating the entire file this way, we still only template `{{content-for 'head'}}` and `{{content-for 'body'}}` in this way to ensure we support other plugins/addons * build: Loosen up the regex for retrieving the CONSUL_VERSION (#5946) * build: Loosen up the regex for retrieving the CONSUL_VERSION 1. Previously the `sed` replacement was searching for the CONSUL_VERSION comment at the start of a line, it no longer does this to allow for indentation. 2. Both `grep` and `sed` where looking for the omment at the end of the line. We've removed this restriction here. We don't need to remove it right now, but if we ever put the comment followed by something here the searching would break. 3. Added `xargs` for trimming the resulting version string. We aren't using this already in the rest of the scripts, but we are pretty sure this is available on most systems. * ui: Fix erroneous variable, and also force an ember cache clean on build 1. We referenced a variable incorrectly here, this fixes that. 2. We also made sure that every `make` target clears ember's `tmp` cache to ensure that its not using any caches that have since been edited everytime we call a `make` target. * added docs, fixed encoding * fixed go fmt * Update agent/config/config.go Co-Authored-By: R.B. Boyer <public@richardboyer.net> * Completed Suggestions * run gofmt on http.go * fix testsanitize * fix fullconfig/hcl by setting correct 'want' * ran gofmt on agent/config/runtime_test.go * Update website/source/docs/agent/options.html.md Co-Authored-By: Hans Hasselberg <me@hans.io> * Update website/source/docs/agent/options.html.md Co-Authored-By: kaitlincarter-hc <43049322+kaitlincarter-hc@users.noreply.github.com> * remove contentpath from redirectFS struct
2019-06-26 16:43:30 +00:00
//UIContentPath is a string that sets the external
// path to a string. Default: /ui/
UIContentPath string
2017-10-19 19:26:02 +00:00
// UnixSocketGroup contains the group of the file permissions when
// Consul binds to UNIX sockets.
//
// hcl: unix_sockets { group = string }
2017-10-19 07:47:22 +00:00
UnixSocketGroup string
2017-10-19 19:26:02 +00:00
// UnixSocketMode contains the mode of the file permissions when
// Consul binds to UNIX sockets.
//
// hcl: unix_sockets { mode = string }
UnixSocketMode string
// UnixSocketUser contains the user of the file permissions when
// Consul binds to UNIX sockets.
//
// hcl: unix_sockets { user = string }
UnixSocketUser string
2017-10-19 07:47:22 +00:00
// VerifyIncoming is used to verify the authenticity of incoming
// connections. This means that TCP requests are forbidden, only allowing
// for TLS. TLS connections must match a provided certificate authority.
// This can be used to force client auth.
//
// hcl: verify_incoming = (true|false)
VerifyIncoming bool
// VerifyIncomingHTTPS is used to verify the authenticity of incoming HTTPS
// connections. This means that TCP requests are forbidden, only allowing
// for TLS. TLS connections must match a provided certificate authority.
// This can be used to force client auth.
//
// hcl: verify_incoming_https = (true|false)
VerifyIncomingHTTPS bool
// VerifyIncomingRPC is used to verify the authenticity of incoming RPC
// connections. This means that TCP requests are forbidden, only allowing
// for TLS. TLS connections must match a provided certificate authority.
// This can be used to force client auth.
//
// hcl: verify_incoming_rpc = (true|false)
VerifyIncomingRPC bool
// VerifyOutgoing is used to verify the authenticity of outgoing
// connections. This means that TLS requests are used. TLS connections must
// match a provided certificate authority. This is used to verify
// authenticity of server nodes.
//
// hcl: verify_outgoing = (true|false)
VerifyOutgoing bool
// VerifyServerHostname is used to enable hostname verification of servers.
// This ensures that the certificate presented is valid for
// server.<datacenter>.<domain>. This prevents a compromised client from
// being restarted as a server, and then intercepting request traffic as
// well as being added as a raft peer. This should be enabled by default
// with VerifyOutgoing, but for legacy reasons we cannot break existing
// clients.
//
// hcl: verify_server_hostname = (true|false)
VerifyServerHostname bool
// Watches are used to monitor various endpoints and to invoke a
// handler to act appropriately. These are managed entirely in the
// agent layer using the standard APIs.
//
// See https://www.consul.io/docs/agent/watches.html for details.
//
// hcl: watches = [
// { type=string ... },
// { type=string ... },
// ...
// ]
//
2017-10-19 07:47:22 +00:00
Watches []map[string]interface{}
EnterpriseRuntimeConfig
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
}
func (c *RuntimeConfig) apiAddresses(maxPerType int) (unixAddrs, httpAddrs, httpsAddrs []string) {
if len(c.HTTPSAddrs) > 0 {
for i, addr := range c.HTTPSAddrs {
if maxPerType < 1 || i < maxPerType {
httpsAddrs = append(httpsAddrs, addr.String())
} else {
break
}
}
}
if len(c.HTTPAddrs) > 0 {
unix_count := 0
http_count := 0
for _, addr := range c.HTTPAddrs {
switch addr.(type) {
case *net.UnixAddr:
if maxPerType < 1 || unix_count < maxPerType {
unixAddrs = append(unixAddrs, addr.String())
unix_count += 1
}
default:
if maxPerType < 1 || http_count < maxPerType {
httpAddrs = append(httpAddrs, addr.String())
http_count += 1
}
}
}
}
return
}
func (c *RuntimeConfig) ClientAddress() (unixAddr, httpAddr, httpsAddr string) {
unixAddrs, httpAddrs, httpsAddrs := c.apiAddresses(0)
if len(unixAddrs) > 0 {
unixAddr = "unix://" + unixAddrs[0]
}
http_any := ""
if len(httpAddrs) > 0 {
for _, addr := range httpAddrs {
host, port, err := net.SplitHostPort(addr)
if err != nil {
continue
}
if host == "0.0.0.0" || host == "::" {
if http_any == "" {
if host == "0.0.0.0" {
http_any = net.JoinHostPort("127.0.0.1", port)
} else {
http_any = net.JoinHostPort("::1", port)
}
}
continue
}
httpAddr = addr
break
}
if httpAddr == "" && http_any != "" {
httpAddr = http_any
}
}
https_any := ""
if len(httpsAddrs) > 0 {
for _, addr := range httpsAddrs {
host, port, err := net.SplitHostPort(addr)
if err != nil {
continue
}
if host == "0.0.0.0" || host == "::" {
if https_any == "" {
if host == "0.0.0.0" {
https_any = net.JoinHostPort("127.0.0.1", port)
} else {
https_any = net.JoinHostPort("::1", port)
}
}
continue
}
httpsAddr = addr
break
}
if httpsAddr == "" && https_any != "" {
httpsAddr = https_any
}
}
return
}
func (c *RuntimeConfig) APIConfig(includeClientCerts bool) (*api.Config, error) {
cfg := &api.Config{
Datacenter: c.Datacenter,
TLSConfig: api.TLSConfig{InsecureSkipVerify: !c.VerifyOutgoing},
}
unixAddr, httpAddr, httpsAddr := c.ClientAddress()
if httpsAddr != "" {
cfg.Address = httpsAddr
cfg.Scheme = "https"
cfg.TLSConfig.CAFile = c.CAFile
cfg.TLSConfig.CAPath = c.CAPath
if includeClientCerts {
cfg.TLSConfig.CertFile = c.CertFile
cfg.TLSConfig.KeyFile = c.KeyFile
}
} else if httpAddr != "" {
cfg.Address = httpAddr
cfg.Scheme = "http"
} else if unixAddr != "" {
cfg.Address = unixAddr
// this should be ignored - however we are still talking http over a unix socket
// so it makes sense to set it like this
cfg.Scheme = "http"
} else {
return nil, fmt.Errorf("No suitable client address can be found")
}
return cfg, nil
}
// Sanitized returns a JSON/HCL compatible representation of the runtime
// configuration where all fields with potential secrets had their
// values replaced by 'hidden'. In addition, network addresses and
// time.Duration values are formatted to improve readability.
func (c *RuntimeConfig) Sanitized() map[string]interface{} {
return sanitize("rt", reflect.ValueOf(c)).Interface().(map[string]interface{})
}
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
func (c *RuntimeConfig) ToTLSUtilConfig() tlsutil.Config {
return tlsutil.Config{
VerifyIncoming: c.VerifyIncoming,
VerifyIncomingRPC: c.VerifyIncomingRPC,
VerifyIncomingHTTPS: c.VerifyIncomingHTTPS,
VerifyOutgoing: c.VerifyOutgoing,
VerifyServerHostname: c.VerifyServerHostname,
CAFile: c.CAFile,
CAPath: c.CAPath,
CertFile: c.CertFile,
KeyFile: c.KeyFile,
NodeName: c.NodeName,
Domain: c.DNSDomain,
ServerName: c.ServerName,
TLSMinVersion: c.TLSMinVersion,
CipherSuites: c.TLSCipherSuites,
PreferServerCipherSuites: c.TLSPreferServerCipherSuites,
EnableAgentTLSForChecks: c.EnableAgentTLSForChecks,
AutoEncryptTLS: c.AutoEncryptTLS,
}
}
// isSecret determines whether a field name represents a field which
// may contain a secret.
func isSecret(name string) bool {
name = strings.ToLower(name)
return strings.Contains(name, "key") || strings.Contains(name, "token") || strings.Contains(name, "secret")
}
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
// cleanRetryJoin sanitizes the go-discover config strings key=val key=val...
// by scrubbing the individual key=val combinations.
func cleanRetryJoin(a string) string {
var fields []string
for _, f := range strings.Fields(a) {
if isSecret(f) {
kv := strings.SplitN(f, "=", 2)
fields = append(fields, kv[0]+"=hidden")
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
} else {
fields = append(fields, f)
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
}
}
return strings.Join(fields, " ")
}
func sanitize(name string, v reflect.Value) reflect.Value {
typ := v.Type()
switch {
// check before isStruct and isPtr
case isNetAddr(typ):
if v.IsNil() {
return reflect.ValueOf("")
}
switch x := v.Interface().(type) {
case *net.TCPAddr:
return reflect.ValueOf("tcp://" + x.String())
case *net.UDPAddr:
return reflect.ValueOf("udp://" + x.String())
case *net.UnixAddr:
return reflect.ValueOf("unix://" + x.String())
case *net.IPAddr:
return reflect.ValueOf(x.IP.String())
case *net.IPNet:
return reflect.ValueOf(x.String())
default:
return v
}
// check before isNumber
case isDuration(typ):
x := v.Interface().(time.Duration)
return reflect.ValueOf(x.String())
case isString(typ):
if strings.HasPrefix(name, "RetryJoinLAN[") || strings.HasPrefix(name, "RetryJoinWAN[") {
x := v.Interface().(string)
return reflect.ValueOf(cleanRetryJoin(x))
}
if isSecret(name) {
return reflect.ValueOf("hidden")
}
return v
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
case isNumber(typ) || isBool(typ):
return v
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
case isPtr(typ):
if v.IsNil() {
return v
}
return sanitize(name, v.Elem())
case isStruct(typ):
m := map[string]interface{}{}
for i := 0; i < typ.NumField(); i++ {
key := typ.Field(i).Name
m[key] = sanitize(key, v.Field(i)).Interface()
}
return reflect.ValueOf(m)
case isArray(typ) || isSlice(typ):
ma := make([]interface{}, 0, v.Len())
for i := 0; i < v.Len(); i++ {
ma = append(ma, sanitize(fmt.Sprintf("%s[%d]", name, i), v.Index(i)).Interface())
}
return reflect.ValueOf(ma)
case isMap(typ):
m := map[string]interface{}{}
for _, k := range v.MapKeys() {
key := k.String()
m[key] = sanitize(key, v.MapIndex(k)).Interface()
}
return reflect.ValueOf(m)
default:
return v
}
}
func isDuration(t reflect.Type) bool { return t == reflect.TypeOf(time.Second) }
func isMap(t reflect.Type) bool { return t.Kind() == reflect.Map }
func isNetAddr(t reflect.Type) bool { return t.Implements(reflect.TypeOf((*net.Addr)(nil)).Elem()) }
func isPtr(t reflect.Type) bool { return t.Kind() == reflect.Ptr }
func isArray(t reflect.Type) bool { return t.Kind() == reflect.Array }
func isSlice(t reflect.Type) bool { return t.Kind() == reflect.Slice }
func isString(t reflect.Type) bool { return t.Kind() == reflect.String }
func isStruct(t reflect.Type) bool { return t.Kind() == reflect.Struct }
func isBool(t reflect.Type) bool { return t.Kind() == reflect.Bool }
func isNumber(t reflect.Type) bool { return isInt(t) || isUint(t) || isFloat(t) || isComplex(t) }
func isInt(t reflect.Type) bool {
return t.Kind() == reflect.Int ||
t.Kind() == reflect.Int8 ||
t.Kind() == reflect.Int16 ||
t.Kind() == reflect.Int32 ||
t.Kind() == reflect.Int64
}
func isUint(t reflect.Type) bool {
return t.Kind() == reflect.Uint ||
t.Kind() == reflect.Uint8 ||
t.Kind() == reflect.Uint16 ||
t.Kind() == reflect.Uint32 ||
t.Kind() == reflect.Uint64
}
func isFloat(t reflect.Type) bool { return t.Kind() == reflect.Float32 || t.Kind() == reflect.Float64 }
func isComplex(t reflect.Type) bool {
return t.Kind() == reflect.Complex64 || t.Kind() == reflect.Complex128
New config parser, HCL support, multiple bind addrs (#3480) * new config parser for agent This patch implements a new config parser for the consul agent which makes the following changes to the previous implementation: * add HCL support * all configuration fragments in tests and for default config are expressed as HCL fragments * HCL fragments can be provided on the command line so that they can eventually replace the command line flags. * HCL/JSON fragments are parsed into a temporary Config structure which can be merged using reflection (all values are pointers). The existing merge logic of overwrite for values and append for slices has been preserved. * A single builder process generates a typed runtime configuration for the agent. The new implementation is more strict and fails in the builder process if no valid runtime configuration can be generated. Therefore, additional validations in other parts of the code should be removed. The builder also pre-computes all required network addresses so that no address/port magic should be required where the configuration is used and should therefore be removed. * Upgrade github.com/hashicorp/hcl to support int64 * improve error messages * fix directory permission test * Fix rtt test * Fix ForceLeave test * Skip performance test for now until we know what to do * Update github.com/hashicorp/memberlist to update log prefix * Make memberlist use the default logger * improve config error handling * do not fail on non-existing data-dir * experiment with non-uniform timeouts to get a handle on stalled leader elections * Run tests for packages separately to eliminate the spurious port conflicts * refactor private address detection and unify approach for ipv4 and ipv6. Fixes #2825 * do not allow unix sockets for DNS * improve bind and advertise addr error handling * go through builder using test coverage * minimal update to the docs * more coverage tests fixed * more tests * fix makefile * cleanup * fix port conflicts with external port server 'porter' * stop test server on error * do not run api test that change global ENV concurrently with the other tests * Run remaining api tests concurrently * no need for retry with the port number service * monkey patch race condition in go-sockaddr until we understand why that fails * monkey patch hcl decoder race condidtion until we understand why that fails * monkey patch spurious errors in strings.EqualFold from here * add test for hcl decoder race condition. Run with go test -parallel 128 * Increase timeout again * cleanup * don't log port allocations by default * use base command arg parsing to format help output properly * handle -dc deprecation case in Build * switch autopilot.max_trailing_logs to int * remove duplicate test case * remove unused methods * remove comments about flag/config value inconsistencies * switch got and want around since the error message was misleading. * Removes a stray debug log. * Removes a stray newline in imports. * Fixes TestACL_Version8. * Runs go fmt. * Adds a default case for unknown address types. * Reoders and reformats some imports. * Adds some comments and fixes typos. * Reorders imports. * add unix socket support for dns later * drop all deprecated flags and arguments * fix wrong field name * remove stray node-id file * drop unnecessary patch section in test * drop duplicate test * add test for LeaveOnTerm and SkipLeaveOnInt in client mode * drop "bla" and add clarifying comment for the test * split up tests to support enterprise/non-enterprise tests * drop raft multiplier and derive values during build phase * sanitize runtime config reflectively and add test * detect invalid config fields * fix tests with invalid config fields * use different values for wan sanitiziation test * drop recursor in favor of recursors * allow dns_config.udp_answer_limit to be zero * make sure tests run on machines with multiple ips * Fix failing tests in a few more places by providing a bind address in the test * Gets rid of skipped TestAgent_CheckPerformanceSettings and adds case for builder. * Add porter to server_test.go to make tests there less flaky * go fmt
2017-09-25 18:40:42 +00:00
}