[SOLVED] Setting unexisting node to 'maintenance mode' adds it to HA manager

[lodex]

Hi there!
I was trying to set a node in our cluster to maintenance mode, but had a typo in the node name, so instead of typing ha-manager crm-command node-maintenance enable pve11 the node name was pve01. To my surprise there was no error message displayed and after taking a look at the HA Status in the GUI, a new node with pve01 as the name was displayed:


Aside from that being a bug in ha-manager, how do I get rid off this "new" node?

 

[t.lamprecht]

Hi,

thanks for your report! As setting maintenance mode requires quorum anyway we can also bail out if the node seems unknown just fine.

Fixed in git will be included in the next pve-ha-manager version, once bumped.

 

[t.lamprecht]

Aside from that being a bug in ha-manager, how do I get rid off this "new" node?

On all nodes: systemctl stop pve-ha-crm.service

On one node: echo '{}' > /etc/pve/ha/manager_status

Then again on all nodes: systemctl start pve-ha-crm.service

 

[lodex]

Thanks a lot for the quick reply. Good to know its already been addressed.

One last question: Since this is a productive cluster, will stopping the CRM service cause any interference with the running VMs?

 

[t.lamprecht]

One last question: Since this is a productive cluster, will stopping the CRM service cause any interference with the running VMs?

Very unlikely as long as the cluster is working and stable - stopping and starting the CRM itself has normally fewer implications and less work to do than e.g., stopping & starting the LRM.

But it might make sense to stop the CRM current active master last; as otherwise a CRM from another node might already have gotten them self elected as new active leader.