[DRBD-user] crm-fence-peer / crm-unfence-peer never terminates

[Felix Zachlod]

Hello group,

I have configured the following:


         handlers {
                 fence-peer              "/usr/lib/drbd/crm-fence-peer.sh";
                 after-resync-target 
"/usr/lib/drbd/crm-unfence-peer.sh";
         }


As well as

         disk {
                 fencing resource-and-stonith;
         }


While the fencing and unfencing in case of the lost replication link 
works as expected - the resource is being demoted / promoted on one side 
and the data is marked outdated, it always leaves a process behind


29522 ?        S      0:00 /bin/bash /usr/lib/drbd/crm-fence-peer.sh
29523 ?        S      0:00  \_ logger -t crm-fence-peer.sh[29521] -p 
local5.info

The script seems actually to be waiting for the sub process to 
terminate, which does not as the logger command which is invoked here 
waits for data on stdin and actually nothing is logged. As soon as I 
kill the logger process the parent script is terminating too.

I can observe this in every server with different versions of the 
cluster stack always running drbd 8.4.5 and drbd-utils 8.9.1. Every stop 
and start of a resource leave a crm-unfence-peer.sh process behind and 
every fencing action leaves a crm-fence-peer.sh process behind.

I have tried this running pacemaker 1.1.7, 1.1.10 and 1.1.12 and always 
see the same behaviour.

What am I doing wrong here?

Thanks in advance,
Felix