0017461: systemd-239-30 package prevents some units from stopping

0017461: systemd-239-30 package prevents some units from stopping

The postgresql unit installed in the postgresql:9.6 package on centos 8.2 fails to stop when using systemd-239-30 and running in an LXD container. This issue does not appear with systemd-239-18 and was introduced with the newest package release. We’re having this issue with the Centos 8 LXD image.<br />
<br />
Some other systemd services that we write and install in /etc/systemd/system/ also fail, but I mention postgres here because it’s an easily reproducible case.<br />
<br />
There were 256 patches in the package for systemd-239-18 and 360 in systemd-239-30, so I assume the issue was introduced in a patch between 257 and 360.<br />
<br />
To reproduce: install postgres, initialize the database, start postgresql. After this, postgresql will not stop. The state of the systemd service is stuck in ‘deactivating’ (final-sigkill), even though all processes started by systemd (postmaster, postgresql-check-db-dir, and postgres itself) are dead according to systemd and according to `ps`.

* This article was originally published here

Leave a Reply

Your email address will not be published. Required fields are marked *