Re: BUG #7670: BUG #7545: Unresponsive server with error log reporting: "poll() failed: Invalid argument"

On 2012-11-18 19:22:04 +0200, Igor wrote:
> 2012/11/18 Andres Freund <andres@2ndquadrant.com>
>
> > On 2012-11-18 18:53:13 +0200, Igor wrote:
> > >  ..
> > >
> > > > Could you give a bit more context about when exactly that error is
> > > > happening?
> > > > It would also be helpful if you could set log_error_verbosity =
> > > > 'verbose'; in the config file and report back with the complete error
> > > > message, including line numbers.
> > > >
> > > >
> > > I try to restart DB and seems get this situation:
> > >
> > > ===
> > > 2012-11-18 18:30:25 EET    pid:93744  LOG:  received SIGHUP, reloading
> > > configuration files
> > > 2012-11-18 18:30:25 EET    pid:93744  LOG:  00000: parameter
> > > "log_error_verbosity" changed to "verbose"
> > > 2012-11-18 18:30:25 EET    pid:93744  LOCATION:  ProcessConfigFile,
> > > guc-file.l:318
> > > 2012-11-18 18:36:45 EET    pid:93744  LOG:  00000: received fast shutdown
> > > request
> > > 2012-11-18 18:36:45 EET    pid:93744  LOCATION:  pmdie, postmaster.c:2187
> > > 2012-11-18 18:36:45 EET    pid:93744  LOG:  00000: aborting any active
> > > transactions
> > > 2012-11-18 18:36:45 EET    pid:93744  LOCATION:  pmdie, postmaster.c:2211
> > > 2012-11-18 18:36:45 EET    pid:93749  LOG:  00000: autovacuum launcher
> > > shutting down
> > > 2012-11-18 18:36:45 EET    pid:93749  LOCATION:  AutoVacLauncherMain,
> > > autovacuum.c:767
> > > 2012-11-18 18:36:45 EET    pid:93746  LOG:  00000: shutting down
> > > 2012-11-18 18:36:45 EET    pid:93746  LOCATION:  ShutdownXLOG,
> > xlog.c:7560
> > > 2012-11-18 18:36:45 EET    pid:93746  LOG:  00000: database system is
> > shut
> > > down
> > > 2012-11-18 18:36:45 EET    pid:93746  LOCATION:  ShutdownXLOG,
> > xlog.c:7582
> > > 2012-11-18 18:36:46 EET    pid:95528  FATAL:  XX000: poll() failed:
> > Invalid
> > > argument
> > > 2012-11-18 18:36:46 EET    pid:95528  LOCATION:  WaitLatchOrSocket,
> > > pg_latch.c:296
> > > ===
> >
> > Ok, this is already helpful. Do you have anything the log that shows
> > what process pid 95528 is? Could you grep for it?
> >
>
> No, this string - is the last string in the log for this event.

Actually I would be interested in *earlier* entries in the log ;)


> > Are you able to recompile postgres with other compilation options and/or
> > some debugging patches applied?

> Will try. Give me this patches.

Ok, let me prepare something.

> > The first thing would be to recompile postgres with --enable-cassert...
> >
>
> That all ?

Yes, that could already catch the source of the error. Note that you
probably will want to revert to non --enable-cassert builds after we
debugged this, it has a non-negligible overhead.

> Do you understand my english ? :)

I try ;)

Greetings,

Andres Freund

--
 Andres Freund                       http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services