[MXS-2321] It should not be possible to put the master into maintenance mode Created: 2019-02-07  Updated: 2020-03-27  Resolved: 2020-03-02

Status: Closed
Project: MariaDB MaxScale
Component/s: mariadbmon, Monitor
Affects Version/s: None
Fix Version/s: 2.5.0

Type: Bug Priority: Major
Reporter: Johan Wikman Assignee: Johan Wikman
Resolution: Fixed Votes: 0
Labels: None

Epic Link: Failover / Recovery Improvements
Sprint: MXS-SPRINT-84, MXS-SPRINT-98, MXS-SPRINT-99, MXS-SPRINT-100

 Description   

Currently it is possible to put the master in maintenance mode, but that will not trigger a failover.

Either,

  • it should not be possible to put the master in maintenance mode, but a switchover would have to be performed first, or
  • if the master is put into maintenance mode, then that should trigger a failover.

The first alternative is better, as that does not lead to implicit actions. Essentially, in the case of a Master/Slave cluster it should only be possible to put a slave in maintenance mode.



 Comments   
Comment by Johan Wikman [ 2019-06-11 ]

This is a bit more complex than what appears on the surface.

While it would make sense to prevent setting the master of a master/slave cluster to maintenance mode, it makes no sense in the context of Clustrix where every node is currently considered a master.

This needs to be done so that the monitor is involved; e.g. in the case of a M/S cluster (defined by the monitor) the setting of the master to maintenance mode should be prevented, but in the case of a Clustrix cluster, the setting of a server to maintenance mode should be allowed until there is only one node left.

Generated at Thu Feb 08 04:13:17 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.