If the pod holding the upgrade lock crashes and is unable to release the lock, then it should expire automatically after some --lock-expiry=6h period, allowing other nodes to continue.
Currently, if the pod restarts after crashing, it will keep the lock acquired and retry the upgrade (once scheduled)... Might be better to ensure any running host-upgrades.service is stopped after restarting, and release the lock instead?
If the pod holding the upgrade lock crashes and is unable to release the lock, then it should expire automatically after some
--lock-expiry=6h
period, allowing other nodes to continue.Currently, if the pod restarts after crashing, it will keep the lock acquired and retry the upgrade (once scheduled)... Might be better to ensure any running
host-upgrades.service
is stopped after restarting, and release the lock instead?