BUG #14851: Systemd kills long-running recovery

On Fri, Oct 13, 2017 at 6:10 AM, <gdr@gdr.name> wrote:

The following bug has been logged on the website:

Bug reference: 14851
Logged by: GDR !
Email address: gdr@gdr.name
PostgreSQL version: 9.6.5
Operating system: CentOS 7 / pgdg
Description:

Summary: I think the line "TimeoutSec=300" in the systemd unit file isn't
necessary

Problem: I'm running a Postgres replication slave fresh after
pg_basebackup. It has about 12 hours worth of xlogs to replay. After 5
minutes of "systemctl start postgresql-9.6" not returning to shell,

systemd

kills the server in the middle of recovery.

This is because of TimeoutSec=300 in the systemd unit file. I don't think
removing this statement will have any negative effects and I propose that
it's not included in the pgdg distribution.

Hi,

I don't think here is the best place to report this issue. Try report it to
the RPM Build maintainers [1]https://yum.postgresql.org/contact.php.

Regards,

[1]: https://yum.postgresql.org/contact.php

--
Fabrízio de Royes Mello
Consultoria/Coaching PostgreSQL

On 10/13/2017 2:10 AM, gdr@gdr.name wrote:

Summary: I think the line "TimeoutSec=300" in the systemd unit file isn't
necessary

Problem: I'm running a Postgres replication slave fresh after
pg_basebackup. It has about 12 hours worth of xlogs to replay. After 5
minutes of "systemctl start postgresql-9.6" not returning to shell, systemd
kills the server in the middle of recovery.

This is because of TimeoutSec=300 in the systemd unit file. I don't think
removing this statement will have any negative effects and I propose that
it's not included in the pgdg distribution.

Maybe the unit code that detects the server is running needs some way of
detecting its in recovery, and that the recovery is somehow progressing,
and report that as 'running' ?

--
john r pierce, recycling bits in santa cruz

--
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs

On October 13, 2017 9:03:57 PM GMT+02:00, John R Pierce <pierce@hogranch.com> wrote:

Maybe the unit code that detects the server is running needs some way
of
detecting its in recovery, and that the recovery is somehow
progressing,
and report that as 'running' ?

John,

As Fabrizio suggested, I have created a bug report at the packaging redmine:

https://redmine.postgresql.org/issues/2786

I'm not following Postgres bugs and I'm only an user since 9.4 so my experience is moderate, but I've not yet encountered a scenario where automatically killing a database server is desirable when it's slow to start up. Do you have some scenario in mind where you'd like systemd to kill a server but only when it's not in recovery?

--
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs