[Pgpool-hackers] pgbench causes pgpool to emit segmentation fault

[Tatsuo Ishii]

> Hi,
> 
> When I run the pgbench to pgpool-II with one master and one standby
> (in streaming replication mode), I got the segmentation fault and pgbench
> got stuck infinitely. Here is the debug log of pgpool.
> 
> 2011-01-31 18:21:04 DEBUG: pid 26567: pool_ssl: SSL requested but SSL
> support is not available
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: auth kind: 0
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: backend key data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: transaction state: I
> 2011-01-31 18:21:04 DEBUG: pid 26567: pool_ssl: SSL requested but SSL
> support is not available
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: auth kind: 0
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: parameter status data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: backend key data received
> 2011-01-31 18:21:04 DEBUG: pid 26567: s_do_auth: transaction state: I
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: T
> 2011-01-31 18:21:04 DEBUG: pid 26567: num_fileds: 1
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: D
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: C
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: Z
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: T
> 2011-01-31 18:21:04 DEBUG: pid 26567: num_fileds: 1
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: D
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: C
> 2011-01-31 18:21:04 DEBUG: pid 26567: do_query: kind: Z
> 2011-01-31 18:21:04 DEBUG: pid 26563: reap_handler called
> 2011-01-31 18:21:04 DEBUG: pid 26563: reap_handler: call wait3
> 2011-01-31 18:21:04 ERROR: pid 26563: Child process 26567 was
> terminated by segmentation fault
> 2011-01-31 18:21:04 LOG:   pid 26563: worker child 26567 exits with
> status 11 by signal 11
> 2011-01-31 18:21:04 DEBUG: pid 26840: I am 26840
> 2011-01-31 18:21:04 LOG:   pid 26563: fork a new worker child pid 26840
> 2011-01-31 18:21:04 DEBUG: pid 26563: reap_handler: normally exited
> 2011-01-31 18:21:17 DEBUG: pid 26565: child received shutdown request signal 2
> 2011-01-31 18:21:17 DEBUG: pid 26563: shmem_exit(0)
> 
> In my machine, this happens when I use several dozen connections in
> pgbench as follows.
> 
>     $ pgbench -c 30 -T1000
>     (scale factor is 1)
> 
> Is this a bug?

Sound like a bug. I will try to reproduce the bug. If I cann't, I will
request you the stack trace.

> I installed pgpool-II from the CVS HEAD, and PostgreSQL from the current
> git master.

Just out of curiousity, have you tried with PostgreSQL 9.0.x? If so,
is there any difference?

> Regards,
> 
> -- 
> Fujii Masao
> NIPPON TELEGRAPH AND TELEPHONE CORPORATION
> NTT Open Source Software Center
> _______________________________________________
> Pgpool-hackers mailing list
> Pgpool-hackers at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgpool-hackers