[pgpool-general: 8374] Re: pgpool failing over on bad queries

[Tatsuo Ishii]

>>Syntax error itself does not trigger failover. There must be other
>>reason which triggers the failover. To find the cause, I need to
>>inspect the pgpool log file.
>>
> What should I look for in the pgpool log file?

When a failover is requested, there will be a log entry something like:

LOG:  received degenerate backend request for node_id: 0 from pid...

> I have logs of log messages "new IPC connection received", those appear normal.
> I also see "child process with pid ... exits with status 256" and "fork a new child process..."

They are normal.

> I see the following on psql-03 when the primary node changed:
> 
> 2022-08-31 01:11:57.819: watchdog pid 971756: LOG:  watchdog received the failover command from remote pgpool-II node "psql-01:9898 Linux psql-01"

I think psql-03 is the main watchdog node and it received a failover
request from pgsql-01 watchdog. The real cause of the failover request
is shown in the pgsql-01 log.

> On psql-01 I see
> 2022-08-31 01:11:57.818: sr_check_worker pid 2111409: LOG:  pool_ssl: "SSL_read": "no SSL error reported"
> 2022-08-31 01:11:57.818: sr_check_worker pid 2111409: LOG:  received degenerate backend request for node_id: 0 from pid [2111409]

The process which originally requested a failover was streaming
replication check process. The error message:

pool_ssl: "SSL_read": "no SSL error reported"

looks strange to me because SSL reported no errors. At this point I
have no idea why this could happen.  Possible orkaround is, disabling
SSL for streaming replication check process.  To do so, change
pg_hba.conf so that the user specified by sr_check_user uses non-SSL
connection.

Best reagards,
--
Tatsuo Ishii
SRA OSS LLC
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp