Configuration

You have to define the etcd cluster in the config inventory before deploying it.

Usually you can choose an etcd cluster with:

• One Node, no high availability, just the functionally of etcd, suitable for development, testing, demonstration.
• Three Nodes, basic high availability, tolerate one node failure, suitable for medium production environment.
• Five Nodes, better high availability, tolerate two node failure, suitable for large production environment.

Even number of etcd nodes is meaningless, and more than five nodes is not common.

One Node

Define the group etcd in the inventory, It will create a singleton etcd instance.

# etcd cluster for ha postgres
etcd: { hosts: { 10.10.10.10: { etcd_seq: 1 } }, vars: { etcd_cluster: etcd } }

This line exists almost in all single-node config templates, where the placeholder IP address 10.10.10.10 will be replaced with the current admin node IP.

The only necessary parameters are etcd_seq and etcd_cluster, which uniquely identify the cluster and each instance.

Three Nodes

It’s common to use a three-node etcd cluster, which can tolerate one node failure, suitable for medium prod env.

The trio and safe config templates use a three-node etcd cluster, as shown below:

etcd: # dcs service for postgres/patroni ha consensus
  hosts:  # 1 node for testing, 3 or 5 for production
    10.10.10.10: { etcd_seq: 1 }  # etcd_seq required
    10.10.10.11: { etcd_seq: 2 }  # assign from 1 ~ n
    10.10.10.12: { etcd_seq: 3 }  # odd number please
  vars: # cluster level parameter override roles/etcd
    etcd_cluster: etcd  # mark etcd cluster name etcd
    etcd_safeguard: false # safeguard against purging
    etcd_clean: true # purge etcd during init process

Five Nodes

Five nodes etcd cluster can tolerate two node failure, suitable for large prod env.

There’s a five-node etcd cluster example in the prod template:

etcd:
  hosts:
    10.10.10.21 : { etcd_seq: 1 }
    10.10.10.22 : { etcd_seq: 2 }
    10.10.10.23 : { etcd_seq: 3 }
    10.10.10.24 : { etcd_seq: 4 }
    10.10.10.25 : { etcd_seq: 5 }
  vars: { etcd_cluster: etcd    }

You can use more nodes for production environment, but 3 or 5 nodes are recommended. Remember to use odd number for cluster size.

Etcd Usage

These are the services that currently use Etcd:

• patroni: Use etcd as a consensus backend for PostgreSQL HA
• vip-manager: Read leader info from Etcd to bind an optional L2 VIP on the PostgreSQL cluster

You’ll have to reload etcd config after any permanent change to the etcd cluster members.

e.g, update patroni reference to etcd endpoints:

./pgsql.yml -t pg_conf                            # generate patroni config
ansible all -f 1 -b -a 'systemctl reload patroni' # reload patroni config

e.g, update vip-manager reference to etcd endpoints (if you are using PGSQL L2 VIP):

./pgsql.yml -t pg_vip_config                           # generate vip-manager config
ansible all -f 1 -b -a 'systemctl restart vip-manager' # restart vip-manager to use