- Parallels Cloud Server version installed is PCS 6.0 update 5.
- High Availability is configured on a Cloud Storage cluster in accordance with the documentation.
- Containers or Virtual Machines were joined to cluster with the help of
- On a node crash (or any other kind of unavailability), the resources are not getting relocated to another node.
- The corresponding CT/VM resource gets marked as 'Broken' in
/var/log/shaman.logcontains no errors about the failed relocation.
The issue is recognized as a product bug with internal ID PSBM-24950. Virtual Environments, added to cluster via
prlctl tool with
--ha-enable key have a missing
PATH variable inside their shaman configuration file, e.g.:
[root@pcs1 ~]# cat /pstorage/pcs2/.shaman/md.569f08fb2ae44b9c/resources/ct-104 RELOCATE_PRIO=0 LAST_NODE=569f08fb2ae44b9c
The correct configuration example:
[root@pcs1 ~]# cat /pstorage/pcs2/.shaman/md.569f08fb2ae44b9c/resources/ct-104 RELOCATE_PRIO=0 PATH=private/104 LAST_NODE=569f08fb2ae44b9c
Rejoin the node into HA cluster:
# service shamand stop # shaman leave # shaman -c $CLUSTERNAME join # service shamand start
PATH variable can be added manually into the corresponding file:
# echo 'PATH=private/104' >> /pstorage/pcs2/.shaman/md.569f08fb2ae44b9c/resources/ct-104