» Multiple Datacenters

» Basic Federation with the WAN Gossip Pool

One of the key features of Consul is its support for multiple datacenters. The architecture of Consul is designed to promote a low coupling of datacenters so that connectivity issues or failure of any datacenter does not impact the availability of Consul in other datacenters. This means each datacenter runs independently, each having a dedicated group of servers and a private LAN gossip pool.

In general, data is not replicated between different Consul datacenters. When a request is made for a resource in another datacenter, the local Consul servers forward an RPC request to the remote Consul servers for that resource and return the results. If the remote datacenter is not available, then those resources will also not be available, but that won't otherwise affect the local datacenter. There are some special situations where a limited subset of data can be replicated, such as with Consul's built-in ACL replication capability, or external tools like consul-replicate.

This guide covers the basic form of federating Consul clusters using a single WAN gossip pool, interconnecting all Consul servers. Consul Enterprise version 0.8.0 added support for an advanced multiple datacenter capability. Please see the Advanced Federation Guide for more details.

» Getting Started

To get started, follow the bootstrapping guide to start each datacenter. After bootstrapping, we should have two datacenters now which we can refer to as dc1 and dc2. Note that datacenter names are opaque to Consul; they are simply labels that help human operators reason about the Consul clusters.

To query the known WAN nodes, we use the members command with the -wan parameter:

$ consul members -wan

This will provide a list of all known members in the WAN gossip pool. This should only contain server nodes. Client nodes send requests to a datacenter-local server, so they do not participate in WAN gossip. Client requests are forwarded by local servers to a server in the target datacenter as necessary.

The next step is to ensure that all the server nodes join the WAN gossip pool (include all the servers in all the datacenters):

$ consul join -wan <server 1> <server 2> ...

The join command is used with the -wan flag to indicate we are attempting to join a server in the WAN gossip pool. As with LAN gossip, you only need to join a single existing member, and the gossip protocol will be used to exchange information about all known members. For the initial setup, however, each server will only know about itself and must be added to the cluster. Consul 0.8.0 added WAN join flooding, so if one Consul server in a datacenter joins the WAN, it will automatically join the other servers in its local datacenter that it knows about via the LAN.

Once the join is complete, the members command can be used to verify that all server nodes gossiping over WAN.

We can also verify that both datacenters are known using the HTTP Catalog API:

$ curl http://localhost:8500/v1/catalog/datacenters
["dc1", "dc2"]

As a simple test, you can try to query the nodes in each datacenter:

$ curl http://localhost:8500/v1/catalog/nodes?dc=dc1
$ curl http://localhost:8500/v1/catalog/nodes?dc=dc2

In order to persist the join information, the following can be added to the consul configuration in each of the server nodes in the cluster. For example, in dc1 server nodes: ... "retry_join_wan":[ "dc2-server-1", ... "dc2-server-N" ], ...

There are a few networking requirements that must be satisfied for this to work. Of course, all server nodes must be able to talk to each other. Otherwise, the gossip protocol as well as RPC forwarding will not work. If service discovery is to be used across datacenters, the network must be able to route traffic between IP addresses across regions as well. Usually, this means that all datacenters must be connected using a VPN or other tunneling mechanism. Consul does not handle VPN or NAT traversal for you.

Note that for RPC forwarding to work the bind address must be accessible from remote nodes. Configuring serf_wan, advertise_wan_addr and translate_wan_addrs can lead to a situation where consul members -wan lists remote nodes but RPC operations fail with one of the following errors:

The most likely cause of these errors is that bind_addr is set to a private address preventing the RPC server from accepting connections across the WAN. Setting bind_addr to a public address (or one that can be routed across the WAN) will resolve this issue. Be aware that exposing the RPC server on a public port should only be done after firewall rules have been established.

The translate_wan_addrs configuration provides a basic address rewriting capability.