Updates break cluster (clustercheck)

MySQL & MariaDB Percona XtraDB Cluster 8.x
1

Hi,

We use NinjaOne to apply updates nightly. Every time an update for XtraDB comes out, it overwrites our clustercheck file which holds the credentials for checking the cluster. This is a major problem as it breaks the cluster making it unusable.

It just happened a second time for us and that is an absolute deal-breaker. So, I’ve set the clustercheck file to immutable with chattr +i though I’m hoping that won’t have other side effects.

I thought about making a copy of the file and changing it’s name, but is there a correct way to address this? It would be nice if the script would pull the credentials from a config file that doesn’t get overwritten.

I’m baffled how this is even a thing given it breaks the database. I saw another thread asking about this going back to CentOS 5, but that thread had no replies to it.

2

Hey @dcbrown73,
For an immediate fix, I would rename the file to something else and update your xinetd/haproxy script to use that new filename instead.
That does sound very frustrating. I would have expected the rpm/apt package to detect that the script has local modifications and thus not override it.
Please open a bug on this at https://jira.percona.com/