Considerations for cluster nodes operating in maintenance mode with SBD Fencing

This document (000020366) is provided subject to the disclaimer at the end of this document.

Environment

SUSE Linux Enterprise Server High Availability Extension 15
SUSE Linux Enterprise Server High Availability Extension 12

Situation

Additional considerations should be taken into account when changes should be applied to cluster nodes that have been put into maintenance mode using:

crm configure property maintenance-mode=true

Resolution

If the cluster has been put globally into maintenance-mode, this means the pacemaker layer will not react to any events that happen to the resources. This may lead to the assumption that the administrator can carry out changes at any layer (e.g., network/storage).

It needs to be planned what tasks exactly need to be carried out and if changes apply to lower components of the cluster
(such as storage or network changes), the safest solution is to stop the cluster services on the affected node completely.

Cause

Only the pacemaker layer operates in maintenance-mode but not the corosync layer, which is responsible for the lower operations of the cluster and controls the sbd (STONITH block device daemon) service. If, for whatever reason, the maintenance tasks carried out by the administrator will cause interruptions to the sbd devices (e.g. storage based maintenance-tasks) or to the network devices responsible for the cluster communication, corosync will respond to these actions and may still fence a node.

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.