Automatic Failover Test with only Two Nodes

[ahussainrc]

• I have setup Postgres Cluster using scheme 1 PostgreSQL High-Availability only
• I am using only two nodes node01 and node02, etcd is also on the same nodes.
• I am also using VIP to have a single IP to connect to Postgres.
• node01 is the leader (master) and node02 is Replica
• To test the automatic failover, I shutdown the node02 with reboot command.
• Automatic failover did happen actually and node02 became master but it only happened when node01 was started and up. Is it because the raft need to have more than half etcd servers to be up and running ?
  If yes, I will have to use 3 servers to have this automatic failover which I dont want.

I have so many environments and having 3 nodes for each environment will consume too much resources. Is there a way to achieve automatic failover with only two nodes ?

[vitabaks]

You can deploy database clusters consisting of two servers, but there must be a DCS (etcd) cluster of at least three servers, here are the details http://thesecretlivesofdata.com/raft/

Create one etcd cluster on three separate small VMs and use it for multiple Postgres clusters. Variables dcs_exists: true and patroni_etcd_hosts

[ahussainrc]

Thanks @vitabaks for your response.

Before I close the discussion, could you please give me an idea how much disk and memory would be consumed on etcd nodes for one postgres cluster.

[vitabaks]

Less than 1 gigabyte, because auto compaction is executed.

[ahussainrc]

@vitabaks
Thank you so much.
I really appreciate your quick help and support.