[pgpool-general: 5893] Re: Pgpool-3.7.1 segmentation fault

Tatsuo Ishii ishii at sraoss.co.jp
Thu Jan 25 07:44:06 JST 2018


> I was working through debugging my problem, my debug logs were full of
> health checks reports. I was experimenting with controlling each one
> individually. I can confidently report that health_check_period and
> health_check_periodN work quite well. :)

Good.

> That sounds like a good idea. You might also consider doing the same for
> the socket files.
> 
> I ran into issues with lingering sockets. I'm on an ubuntu system, running
> upstart.  At some point, I ran stop pgpool. I tried start pgpool a few
> minutes later. I found that it could not start, because the socket files
> had not been removed during shutdown and pgpool would not remove them
> during startup.  My init script removes the socket files from the run
> directory and the state file from the log directory.

Hmm. I thought the socket files are removed when pgpool main process exists.
>From src/main/pgpool_main.c:

	on_proc_exit(FileUnlink, (Datum) un_addr.sun_path);

Maybe the stop script kill -9 the pgpool process?

> I also think it would be great if pgpool exposed the number of zero,
> non-zero, and exits due to signals the children generated. Parsing the log
> for that will be a bit cumbersome.

Sounds a nice idea. Is it ok to add those logs as debug log?

If child_life_time or some other timer for exiting pgpool child, you
would see lots of the log lines. So I hasitate to set the log level to
LOG.

> Thanks for the great work! :)

Thanks!
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp


More information about the pgpool-general mailing list