Re: BUG #14835: Postgres crashed

Thanks Tomas.
The query is a select and it takes 4-5s to run. This does not seem the
problem.
They were queries that were able to run for a long time (years) and
postgres never crashed since the last two updates.

For now there were a few updates on Centos 6.9 (Kernel i.e) so let's see if
this will solve the problem.

I'll keep you posted.

Nicola

2017-09-28 19:14 GMT+02:00 Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>:

>
>
> On 09/28/2017 05:46 PM, nicola(dot)contu(at)gmail(dot)com wrote:
> > The following bug has been logged on the website:
> >
> > Bug reference: 14835
> > Logged by: Nicola Contu
> > Email address: nicola(dot)contu(at)gmail(dot)com
> > PostgreSQL version: 9.6.5
> > Operating system: Centos 6.9
> > Description:
> >
> > Hello,
> > after upgrading to postgres 9.6.5 in the last 3weeks we had the db
> crashing
> > three times for the same reason :
> >
> > SELECT: double free or corruption (!prev): 0x0000000001f4b440 ***
> > ======= Backtrace: =========
> > /lib64/libc.so.6[0x3eee675dee]
> > /lib64/libc.so.6[0x3eee678c80]
> >
> >
> > ======= Memory map: ========
> > 00400000-009ff000 r-xp 00000000 fd:00 6431324
> > /usr/local/pgsql9.6.5/bin/postgres
>
> Yes, this seems like a bug, but it's quite unclear if it's in PostgreSQL
> or due to a bug in some other library.
>
> Please, make sure next time a core dump is generated, so that you can
> inspect it, and generate backtrace for example. You'll probably have to
> do this:
>
> 1) install debuginfo packages (likely using debuginfo-install)
>
> 2) enable core files (you'll probably have to set limits in limits.conf
> and set core_pattern), see [1] for details
>
> It might also be useful to know what query was being executed at the
> time of the crash - that should be logged into the PostgreSQL log.
>
> [1] http://man7.org/linux/man-pages/man5/core.5.html
>
> regards
>
> --
> Tomas Vondra http://www.2ndQuadrant.com
> PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
>