2023.3.x Upgrade Fails Due to Postgres Not Starting

Customer Managed Applies to customer-managed instances of Alation

Problem

The upgrade to 2023.3.1 or a later minor version of 2023.3 fails due to Postgres failing to start and an error in the Postgres logs that is similar to the following:

runuser: warning: cannot change directory to /home/postgres: No such file or directory
postmaster: could not access directory "/var/lib/pgsql/13/data": Permission denied

Location of Postgres logs: Postgres Logs.

The issue affects both standalone and high availability types of Alation instances where the installation followed a custom path and the Postgres component had been installed in a Docker container or where the /etc/passwd file had been modified.

Root Cause

This is a bug. Starting with 2023.3.1, the UID of the postgres user was changed, which causes misalignment with the postgresql user/group on existing instances that are being upgraded.

A fix will be provided in a future Alation version.

Solution

If the upgrade fails on your instance due to this issue, perform the following steps on the Alation server after the upgrade fails.

Important

In case of the high availability pair, perform these steps on both primary and secondary servers.

1. On the Alation server, from the Alation shell, run the following command:

   chown -R postgres:postgres /data1/psql /data2/backup/pgbackrest/
   

2. Restart Alation.

   service alation restart
   

3. Re-attempt the upgrade.