Re: pg_dump crashing - Mailing list pgsql-general

From Michael Paquier
Subject Re: pg_dump crashing
Date
Msg-id CAB7nPqQ=E_e8bT1zP_C-SFKHC8CbL-StSsWO1_RXK31C+gbRBg@mail.gmail.com
Whole thread Raw
In response to pg_dump crashing  (Matthias Schmitt <matthias.schmitt@mmp.lu>)
List pgsql-general
On Tue, Mar 15, 2016 at 4:10 PM, Matthias Schmitt
<matthias.schmitt@mmp.lu> wrote:
> since two weeks I am trying to get PostgreSQL 9.5.1 to run on Debian 8.3. Everything is fine except the daily backup.
Whencalling pg_dump as part of a cron job pg_dump crashes: 
>
> 2016-03-15 01:00:02 CETFATAL:  semctl(23232524, 3, SETVAL, 0) failed: Invalid argument
> 2016-03-15 01:00:02 CETLOG:  server process (PID 22279) exited with exit code 1
> [...]
> 4th  possible solution: I have read here: https://lists.freedesktop.org/archives/systemd-devel/2014-April/018373.html
> that logging into the postgres user in and out might delete semaphore memory, but I cannot reproduce this.

That's scary, and an entrance for many crashes... Perhaps there are
some underlying calls of ipcrm?

> Interesting enough I have another PostgreSQL version running perfectly under Debian 8.3 on a different machine. This
isa PostgreSQL 9.4.1. It uses the same backup mechanism. I hate to consider a downgrade as a possible solution. May be
aninfo, which might be useful: my machine has 64 GB RAM. 
>
> Any ideas where to continue my search?

Hm, and both of them are managed with systemd? Are there differences
in each one's spec file, I find rather bad the fact that there is
something removing the semaphores at OS level all-of-a-sudden.
--
Michael


pgsql-general by date:

Previous
From: John Lumby
Date:
Subject: Re: how to switch old replication Master to new Standby after promoting old Standby - pg_rewind log file missing
Next
From: Michael Paquier
Date:
Subject: Re: how to switch old replication Master to new Standby after promoting old Standby - pg_rewind log file missing