Merge branch 'master' of github.com:hashicorp/consul

This commit is contained in:
Hans Hasselberg 2019-03-21 15:13:03 +00:00
commit 0df516aa52
4 changed files with 104 additions and 47 deletions

View File

@ -56,7 +56,7 @@ The following guides are available:
* [Monitoring Consul with Telegraf](/docs/guides/monitoring-telegraf.html) - This guide demonstrates how to setup Consul for monitoring with Telegraf.
* [Network Segments](/docs/guides/segments.html) - Configuring Consul to support partial LAN connectivity using Network Segments.
* [Network Segments](/docs/guides/network-segments.html) - Configuring Consul to support partial LAN connectivity using Network Segments.
* [Outage Recovery](https://learn.hashicorp.com/consul/day-2-operations/advanced-operations/outage) - This guide covers recovering a cluster that has become unavailable due to server failures.

View File

@ -10,14 +10,9 @@ description: |-
segment.
---
# Network Segments
# Network Segments [Enterprise Only]
## Partial LAN Connectivity with Network Segments
[//]: # ( ~> The network segment functionality described here is available only in )
[//]: # ( [Consul Enterprise](https://www.hashicorp.com/products/consul/) version 0.9.3 and later. )
<%= enterprise_alert :consul %>
~> Note, the network segment functionality described here is available only in [Consul Enterprise](https://www.hashicorp.com/products/consul/) version 0.9.3 and later.
Many advanced Consul users have the need to run clusters with segmented networks, meaning that
not all agents can be in a full mesh. This is usually the result of business policies enforced
@ -25,21 +20,30 @@ via network rules or firewalls. Prior to Consul 0.9.3 this was only possible thr
which for some users is too heavyweight or expensive as it requires running multiple servers per
segment.
This guide will cover the basic configuration for setting up multiple segments, as well as
how to configure a prepared query to limit service discovery to the services in the local agent's
network segment.
To complete this guide you will need to complete the
[Deployment Guide](https://learn.hashicorp.com/consul/advanced/day-1-operations/deployment-guide).
## Partial LAN Connectivity with Network Segments
By default, all Consul agents in one datacenter are part of a shared gossip pool over the LAN;
this means that the partial connectivity caused by segmented networks would cause health flapping
as nodes failed to communicate. In this guide we will cover the Network Segments feature, added
in [Consul Enterprise](https://www.hashicorp.com/products/consul/) version 0.9.3, which allows users
to configure Consul to support this kind of segmented network topology.
This guide will cover the basic configuration for setting up multiple segments, as well as
how to configure a prepared query to limit service discovery to the services in the local agent's
network segment.
## Configuring Network Segments
### Network Segments Overview
All Consul agents are part of the default network segment, `""`, unless a segment is specified in
their configuration. In a standard cluster setup all agents will normally be part of this default
segment and as a result, part of one shared LAN gossip pool. Network segments can be used to break
All Consul agents are part of the default network segment, unless a segment is specified in
their configuration. In a standard cluster setup, all agents will normally be part of this default
segment and as a result, part of one shared LAN gossip pool.
Network segments can be used to break
up the LAN gossip pool into multiple isolated smaller pools by specifying the configuration for segments
on the servers. Each desired segment must be given a name and port, as well as optionally a custom
bind and advertise address for that segment's gossip listener to bind to on the server.
@ -62,15 +66,16 @@ agent within the same segment. If joining to a Consul server, client will need t
port for their segment along with the address of the server when performing the join (for example,
`consul agent -retry-join "consul.domain.internal:1234"`).
## Getting Started
## Setup Network Segments
To get started, follow the [bootstrapping guide](/docs/guides/bootstrapping.html) to
start a server or group of servers, with the following section added to the configuration (you may need to
adjust the bind/advertise addresses for your setup):
### Configure Consul Servers
To get started,
start a server or group of servers, with the following section added to the configuration. Note, you may need to
adjust the bind/advertise addresses for your setup.
```json
{
...
"segments": [
{"name": "alpha", "bind": "{{GetPrivateIP}}", "advertise": "{{GetPrivateIP}}", "port": 8303},
{"name": "beta", "bind": "{{GetPrivateIP}}", "advertise": "{{GetPrivateIP}}", "port": 8304}
@ -78,9 +83,9 @@ adjust the bind/advertise addresses for your setup):
}
```
You should see a log message on the servers for each segment's listener as the agent starts up:
You should see a log message on the servers for each segment's listener as the agent starts up.
```text
```sh
2017/08/30 19:05:13 [INFO] serf: EventMemberJoin: server1.dc1 192.168.0.4
2017/08/30 19:05:13 [INFO] serf: EventMemberJoin: server1 192.168.0.4
2017/08/30 19:05:13 [INFO] consul: Started listener for LAN segment "alpha" on 192.168.0.4:8303
@ -89,38 +94,40 @@ You should see a log message on the servers for each segment's listener as the a
2017/08/30 19:05:13 [INFO] serf: EventMemberJoin: server1 192.168.0.4
```
Running `consul members` should show the server as being part of all segments:
Running `consul members` should show the server as being part of all segments.
```text
```sh
(server1) $ consul members
Node Address Status Type Build Protocol DC Segment
server1 192.168.0.4:8301 alive server 0.9.3+ent 2 dc1 <all>
```
Next, start a client agent in the 'alpha' segment, with `-join` set to the server's segment
address/port for that segment:
### Configure Consul Clients in Different Network Segments
```text
Next, start a client agent in the 'alpha' segment, with `-join` set to the server's segment
address/port for that segment.
```sh
(client1) $ consul agent ... -join 192.168.0.4:8303 -node client1 -segment alpha
```
After the join is successful, we should see the client show up by running the `consul members` command
on the server again:
on the server again.
```text
```sh
(server1) $ consul members
Node Address Status Type Build Protocol DC Segment
server1 192.168.0.4:8301 alive server 0.9.3+ent 2 dc1 <all>
client1 192.168.0.5:8301 alive client 0.9.3+ent 2 dc1 alpha
```
Now join another client in segment 'beta' and run the `consul members` command another time:
Now join another client in segment 'beta' and run the `consul members` command another time.
```text
```sh
(client2) $ consul agent ... -join 192.168.0.4:8304 -node client2 -segment beta
```
```text
```sh
(server1) $ consul members
Node Address Status Type Build Protocol DC Segment
server1 192.168.0.4:8301 alive server 0.9.3+ent 2 dc1 <all>
@ -128,10 +135,12 @@ client1 192.168.0.5:8301 alive client 0.9.3+ent 2 dc1 alpha
client2 192.168.0.6:8301 alive client 0.9.3+ent 2 dc1 beta
```
If we pass the `-segment` flag when running `consul members`, we can limit the view to agents
in a specific segment:
### Filter Segmented Nodes
```text
If we pass the `-segment` flag when running `consul members`, we can limit the view to agents
in a specific segment.
```sh
(server1) $ consul members -segment alpha
Node Address Status Type Build Protocol DC Segment
client1 192.168.0.5:8301 alive client 0.9.3+ent 2 dc1 alpha
@ -139,9 +148,9 @@ server1 192.168.0.4:8303 alive server 0.9.3+ent 2 dc1 alpha
```
Using the `consul catalog nodes` command, we can filter on an internal metadata key,
`consul-network-segment`, which stores the network segment of the node:
`consul-network-segment`, which stores the network segment of the node.
```text
```sh
(server1) $ consul catalog nodes -node-meta consul-network-segment=alpha
Node ID Address DC
client1 4c29819c 192.168.0.5 dc1
@ -150,25 +159,31 @@ client1 4c29819c 192.168.0.5 dc1
With this metadata key, we can construct a [Prepared Query](/api/query.html) that can be used
for DNS to return only services within the same network segment as the local agent.
First, register a service on each of the client nodes:
## Configure a Prepared Query to Limit Service Discovery
```text
### Create Services
First, register a service on each of the client nodes.
```sh
(client1) $ curl \
--request PUT \
--data '{"Name": "redis", "Port": 8000}' \
localhost:8500/v1/agent/service/register
```
```text
```sh
(client2) $ curl \
--request PUT \
--data '{"Name": "redis", "Port": 9000}' \
localhost:8500/v1/agent/service/register
```
Next, write the following to `query.json` and create the query using the HTTP endpoint:
### Create the Prepared Query
```text
Next, write the following to `query.json` and create the query using the HTTP endpoint.
```sh
(server1) $ curl \
--request POST \
--data \
@ -186,12 +201,14 @@ Next, write the following to `query.json` and create the query using the HTTP en
{"ID":"6f49dd24-de9b-0b6c-fd29-525eca069419"}
```
### Test the Segments with DNS Lookups
Now, we can replace any dns lookups of the form `<service>.service.consul` with
`<service>.query.consul` to look up only services within the same network segment:
`<service>.query.consul` to look up only services within the same network segment.
**Client 1:**
```text
```sh
(client1) $ dig @127.0.0.1 -p 8600 redis.query.consul SRV
; <<>> DiG 9.8.3-P1 <<>> @127.0.0.1 -p 8600 redis.query.consul SRV
@ -214,7 +231,7 @@ client1.node.dc1.consul. 0 IN A 192.168.0.5
**Client 2:**
```text
```sh
(client2) $ dig @127.0.0.1 -p 8600 redis.query.consul SRV
; <<>> DiG 9.8.3-P1 <<>> @127.0.0.1 -p 8600 redis.query.consul SRV
@ -234,3 +251,10 @@ redis.query.consul. 0 IN SRV 1 1 9000 client2.node.dc1.consul.
;; ADDITIONAL SECTION:
client2.node.dc1.consul. 0 IN A 192.168.0.6
```
## Summary
In this guide you configured the Consul agents to participate in partial
LAN gossip based on network segments. You then set up a couple services and
a prepared query to test the segments.

View File

@ -36,7 +36,40 @@ EOF
```
-> **Note:** The `stubDomain` can only point to a static IP. If the cluster IP
of the `consul-dns` service changes, then it must be updated to continue
of the `consul-dns` service changes, then it must be updated in the config map to
match the new service IP for this to continue
working. This can happen if the service is deleted and recreated, such as
in full cluster rebuilds.
## CoreDNS Configuration
If you are using CoreDNS instead of kube-dns in your Kubernetes cluster, you will
need to update your existing `coredns` ConfigMap in the `kube-system` namespace to
include a proxy definition for `consul` that points to the cluster IP of the
`consul-dns` service.
```
apiVersion: v1
kind: ConfigMap
metadata:
labels:
addonmanager.kubernetes.io/mode: EnsureExists
name: coredns
namespace: kube-system
data:
Corefile: |
.:53 {
<Existing CoreDNS definition>
}
consul {
errors
cache 30
proxy . <consul-dns service cluster ip>
}
```
-> **Note:** The consul proxy can only point to a static IP. If the cluster IP
of the `consul-dns` service changes, then it must be updated to the new IP to continue
working. This can happen if the service is deleted and recreated, such as
in full cluster rebuilds.

View File

@ -472,7 +472,7 @@
<a href="/docs/guides/monitoring-telegraf.html">Monitoring Consul with Telegraf</a>
</li>
<li<%= sidebar_current("docs-guides-segments") %>>
<a href="/docs/guides/segments.html">Network Segments</a>
<a href="/docs/guides/network-segments.html">Network Segments</a>
</li>
<li<%= sidebar_current("docs-guides-outage") %>>
<a href="https://learn.hashicorp.com/consul/day-2-operations/advanced-operations/outage">Outage Recovery</a>