[pgpool-general: 8657] Re: recovery failed to start in 0 second

[Bo Peng]

Hello,

> I sometimes get the following message when trying to recover a down cluster node:
> 
> pcp_recovery_node -h 10.78.121.1 -p 9898 -U pgpool -n 0
> Password:
> ERROR:  recovery is checking if postmaster is started
> DETAIL:  postmaster on hostname:"catvmtspg02a.domain.net" database:"template1" user:"postgres" failed to start in 0 second
> 
> How can I stop this from happening?  The pcp_recovery_node process is usually successful on a later attempt.

After running remote start of the target node, Pgpool-II will try to connect to the PostgreSQL node
to check if postmaster is started.

First, Pgpool-II will try to connect to database "postgres" 3 times.
If it fails, Pgpool-II will try to connect to "template1" periodically
within recovery_timeout/3 seconds.

It seems you have configured recovery_timeout = 0 and 0 means no wait.

If the startup of PostgreSQL takes time, please increase recovery_timeout.

-- 
Bo Peng <pengbo at sraoss.co.jp>
SRA OSS LLC
https://www.sraoss.co.jp/