[ClusterLabs] Any node rebooting with 8x module stops all nodes

[Oyvind Albrigtsen]

Sounds like you need a systemd drop-in file to either set the module
service to be run Before the pacemaker service, or the pacemaker to
run After it:
https://stackoverflow.com/questions/59842743/what-is-a-drop-in-file-what-is-a-drop-in-directory-how-to-edit-systemd-service


Oyvind Albrigtsen

On 25/04/25 08:45 -0400, jayf at wheeling-nipponsteel.com wrote:
>
>Greeting,
>
>Forgive me if this is the wrong spot to ask, because this involve pacemaker
>and DRBD, I'm not sure which direction to go for help.
>I have a 3 node cluster for running DRBD and VMs.  All are Debian12.  Two
>are "disk" nodes, the third is "diskless"
>
>Debian packs the old 8.x version of the DRBD kernel modules, so I compile
>the 9.2 branch modules from source.
>
>Here is my issue, which I'm sure someone else has run into before.  If the
>kernel has been upgraded, and a node is rebooted, it tries coming up with
>the 8.x driver.
>This of course doesn't work with a 3 node DRBD  cluster.  The problem is
>that it wreaks havoc with the whole cluster and the VMs end up failing and
>are offline.
>
>I would hope that if the rebooted node was a "secondary", that the
>"primary" would be okay to continue with perhaps some error.  However, that
>is not the case.
>
>I have gone so far as write a script that is called by cron on reboot to
>check for the version of the module and if it's 8.x it builds/installs the
>new module.
>But it appears that Pacemaker must come up before this process is done.
>
>Any help is appreciated.
>
>Jay Fluharty
>Network Analyst
>Wheeling-Nippon Steel Inc.
>PO Box 635
>Follansbee, WV 26037
>jayf at wheeling-nipponsteel.com
>1-304-527-4819

>_______________________________________________
>Manage your subscription:
>https://lists.clusterlabs.org/mailman/listinfo/users
>
>ClusterLabs home: https://www.clusterlabs.org/