[pgpool-general: 9182] Re: Pgpool-II 4.4.4 Split-brain situation

[Bo Peng]

Hi,

> Hello Team,
> 
> I Need help or suggestions on the issue below pgpool-II.
> 
> Example:-
> Node1 -- Primary (Postgres & pgpool installed)
> Node2 -- Standby (Postgres & pgpool installed)
> Node3 -- Witness  (Pgpool installed)
> 
> Incident:-
> 
> 1. In the pgpool cluster we have node1 is primary, node2 is secondary, and
> node3 is witness.
> 2. node1 becomes unresponsive then failover is triggered and then node2
> becomes the new primary.
> 3. After 15 minutes node1 became responsive again and node1 is in primary
> state and we have node 2 also in primary so a split brain situation
> happened.
> 4. Node 1 itself is decided as leader/coordinator based on the node 1
> pgpool record and more number of connected standby nodes (based on old
> state of information) , now i am the true leader, it is decided as the true
> leader and asks node 2 to step down.
> 5. In this case for 15 minutes data written to node 2 new primary after
> node 1 responsive data gets started to return on node 1 again so here data
> gets inconsistent.
> 
> Please provide a suggestion in this case.
> 
> NOTE:- We are using only for auto-failover not for auto-failback so
> auto-failback is disabled.

To avoid Split-brain, you can use watchdog by setting "use_watchdog = on".
Are you using watchdog?

-- 
Bo Peng <pengbo at sraoss.co.jp>
SRA OSS K.K.
TEL: 03-5979-2701 FAX: 03-5979-2702
URL: https://www.sraoss.co.jp/