consul/command/connect/expose/expose.go

241 lines
6.9 KiB
Go
Raw Normal View History

2020-06-05 21:54:29 +00:00
package expose
import (
"flag"
"fmt"
"strings"
"github.com/hashicorp/consul/agent"
"github.com/hashicorp/consul/agent/structs"
2020-06-05 21:54:29 +00:00
"github.com/hashicorp/consul/api"
"github.com/hashicorp/consul/command/flags"
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
"github.com/hashicorp/consul/command/intention"
2020-06-05 21:54:29 +00:00
"github.com/mitchellh/cli"
)
func New(ui cli.Ui) *cmd {
c := &cmd{UI: ui}
c.init()
return c
}
type cmd struct {
UI cli.Ui
flags *flag.FlagSet
http *flags.HTTPFlags
help string
// flags
ingressGateway string
service string
port int
protocol string
2020-06-08 23:59:47 +00:00
hosts flags.AppendSliceValue
2020-06-05 21:54:29 +00:00
}
func (c *cmd) init() {
c.flags = flag.NewFlagSet("", flag.ContinueOnError)
c.flags.StringVar(&c.ingressGateway, "ingress-gateway", "",
"(Required) The name of the ingress gateway service to use. A namespace "+
"can optionally be specified as a prefix via the 'namespace/service' format.")
2020-06-05 21:54:29 +00:00
c.flags.StringVar(&c.service, "service", "",
"(Required) The name of destination service to expose. A namespace "+
"can optionally be specified as a prefix via the 'namespace/service' format.")
2020-06-05 21:54:29 +00:00
c.flags.IntVar(&c.port, "port", 0,
"(Required) The listener port to use for the service on the Ingress gateway.")
2020-06-05 21:54:29 +00:00
c.flags.StringVar(&c.protocol, "protocol", "tcp",
"The protocol for the service. Defaults to 'tcp'.")
2020-06-05 21:54:29 +00:00
2020-06-08 23:59:47 +00:00
c.flags.Var(&c.hosts, "host", "Additional DNS hostname to use for routing to this service."+
"Can be specified multiple times.")
2020-06-05 21:54:29 +00:00
c.http = &flags.HTTPFlags{}
flags.Merge(c.flags, c.http.ClientFlags())
flags.Merge(c.flags, c.http.ServerFlags())
c.help = flags.Usage(help, c.flags)
}
func (c *cmd) Run(args []string) int {
if err := c.flags.Parse(args); err != nil {
if err == flag.ErrHelp {
return 0
}
c.UI.Error(fmt.Sprintf("Failed to parse args: %v", err))
return 1
}
// Set up a client.
client, err := c.http.APIClient()
if err != nil {
c.UI.Error(fmt.Sprintf("Error initializing client: %s", err))
return 1
}
// Check for any missing or invalid flag values.
if c.service == "" {
c.UI.Error("A service name must be given via the -service flag.")
return 1
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
svc, svcNamespace, err := intention.ParseIntentionTarget(c.service)
2020-06-05 21:54:29 +00:00
if err != nil {
c.UI.Error(fmt.Sprintf("Invalid service name: %s", err))
return 1
}
if c.ingressGateway == "" {
c.UI.Error("An ingress gateway service must be given via the -ingress-gateway flag.")
return 1
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
gateway, gatewayNamespace, err := intention.ParseIntentionTarget(c.ingressGateway)
2020-06-05 21:54:29 +00:00
if err != nil {
c.UI.Error(fmt.Sprintf("Invalid ingress gateway name: %s", err))
return 1
}
if c.port == 0 {
2020-06-05 21:54:29 +00:00
c.UI.Error("A port must be provided via the -port flag.")
return 1
}
// First get the config entry for the ingress gateway, if it exists. Don't error if it's a 404 as that
// just means we'll need to create a new config entry.
conf, _, err := client.ConfigEntries().Get(api.IngressGateway, gateway, nil)
if err != nil && !strings.Contains(err.Error(), agent.ConfigEntryNotFoundErr) {
c.UI.Error(fmt.Sprintf("Error fetching existing ingress gateway configuration: %s", err))
return 1
}
if conf == nil {
conf = &api.IngressGatewayConfigEntry{
Kind: api.IngressGateway,
Name: gateway,
Namespace: gatewayNamespace,
}
}
// Make sure the flags don't conflict with existing config.
ingressConf, ok := conf.(*api.IngressGatewayConfigEntry)
if !ok {
// This should never happen
c.UI.Error(fmt.Sprintf("Config entry is an invalid type: %T", conf))
return 1
}
listenerIdx := -1
serviceIdx := -1
newService := api.IngressService{
Name: svc,
Namespace: svcNamespace,
Hosts: c.hosts,
}
2020-06-05 21:54:29 +00:00
for i, listener := range ingressConf.Listeners {
// Find the listener for the specified port, if one exists.
if listener.Port != c.port {
continue
2020-06-05 21:54:29 +00:00
}
// Make sure the given protocol matches the existing one.
listenerIdx = i
if listener.Protocol != c.protocol {
c.UI.Error(fmt.Sprintf("Listener on port %d already configured with conflicting protocol %q", listener.Port, listener.Protocol))
return 1
}
// Make sure the service isn't already exposed in this gateway
for j, service := range listener.Services {
if service.Name == svc && namespaceMatch(service.Namespace, svcNamespace) {
serviceIdx = j
c.UI.Output(fmt.Sprintf("Updating service definition for %q on listener with port %d", c.service, listener.Port))
break
2020-06-05 21:54:29 +00:00
}
}
}
// Add a service to the existing listener for the port if one exists, or make a new listener.
if listenerIdx >= 0 {
if serviceIdx >= 0 {
ingressConf.Listeners[listenerIdx].Services[serviceIdx] = newService
} else {
ingressConf.Listeners[listenerIdx].Services = append(ingressConf.Listeners[listenerIdx].Services, newService)
}
2020-06-05 21:54:29 +00:00
} else {
ingressConf.Listeners = append(ingressConf.Listeners, api.IngressListener{
Port: c.port,
Protocol: c.protocol,
Services: []api.IngressService{newService},
2020-06-05 21:54:29 +00:00
})
}
// Write the updated config entry using a check-and-set, so it fails if the entry
// has been changed since we looked it up.
succeeded, _, err := client.ConfigEntries().CAS(ingressConf, ingressConf.GetModifyIndex(), nil)
if err != nil {
c.UI.Error(fmt.Sprintf("Error writing ingress config entry: %v", err))
return 1
}
if !succeeded {
c.UI.Error("Ingress config entry was changed while attempting to update, please try again.")
return 1
2020-06-05 21:54:29 +00:00
}
c.UI.Output(fmt.Sprintf("Successfully updated config entry for ingress service %q", gateway))
2020-06-05 21:54:29 +00:00
// Check for an existing intention.
existing, _, err := client.Connect().IntentionGetExact(c.ingressGateway, c.service, nil)
2020-06-05 21:54:29 +00:00
if err != nil {
c.UI.Error(fmt.Sprintf("Error looking up existing intention: %s", err))
return 1
}
if existing != nil && existing.Action == api.IntentionActionAllow {
c.UI.Output(fmt.Sprintf("Intention already exists for %q -> %q", c.ingressGateway, c.service))
2020-06-05 21:54:29 +00:00
return 0
}
// Add the intention between the gateway service and the destination.
ixn := &api.Intention{
SourceName: gateway,
SourceNS: gatewayNamespace,
DestinationName: svc,
DestinationNS: svcNamespace,
SourceType: api.IntentionSourceConsul,
Action: api.IntentionActionAllow,
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
if _, err = client.Connect().IntentionUpsert(ixn, nil); err != nil {
c.UI.Error(fmt.Sprintf("Error upserting intention: %s", err))
return 1
2020-06-05 21:54:29 +00:00
}
c.UI.Output(fmt.Sprintf("Successfully set up intention for %q -> %q", c.ingressGateway, c.service))
return 0
}
func namespaceMatch(a, b string) bool {
namespaceA := a
namespaceB := b
if namespaceA == "" {
namespaceA = structs.IntentionDefaultNamespace
}
if namespaceB == "" {
namespaceB = structs.IntentionDefaultNamespace
}
return namespaceA == namespaceB
}
2020-06-05 21:54:29 +00:00
func (c *cmd) Synopsis() string {
return synopsis
}
func (c *cmd) Help() string {
return c.help
}
const synopsis = "Expose a Connect-enabled service through an Ingress gateway"
const help = `
Usage: consul connect expose [options]
Exposes a Connect-enabled service through the given ingress gateway, using the
given protocol and port.
`