[pgpool-general: 3359] Re: [PATCH] pgpool2 parent process doesn't finish if init_system_db_connection fails

[Sadayuki Furuhashi]

Sorry, I guess I failed to attach the patch.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: system_db_connect_error.patch.txt
URL: <http://www.pgpool.net/pipermail/pgpool-general/attachments/20141210/30943df4/attachment.txt>
-------------- next part --------------


--
Sadayuki Furuhashi
http://fluentd.org http://msgpack.org
twitter:@frsyuki

2014/12/10 17:18, Sadayuki Furuhashi <frsyuki at gmail.com>:

> Hi,
> 
> I found that init_system_db_connection() calls ereport(ERROR) before sigsetjmp().
> An influence of this issue is that child processes won't exit by themselves when init_system_db_connection()
> fails and sending SIGTERM to the parent process doesn't complete shutdown process appropriately.
> 
> This patch replaces ereport(ERROR) with ereport(WARNING) + child_exit(1).
> A trade-off is that pgpool2 shows a lot of "failed to make persistent system db connection" message to the log
> because a child process exists immediately and the parent process keeps restarting them immediately.
> 
> --
> Sadayuki Furuhashi
> http://fluentd.org http://msgpack.org
> twitter:@frsyuki
> 
> <system_db_connect_error.patch>