There have been several issues where an NFS server that is undergoing maintenance or during failovers has issues with locks staying open requiring a reboot of an NFS server during failover (like in the parent task), a non-active server that is used via a symlink causing anything once connected to blow up with load (even if they weren't using any files and are now unmounted) like in T196651 where whenever we shut one down, regardless of use, load skyrocketed throughout toolforge.
We need to, in a relatively safe environment, go over our options for how we can reduce downtime when an NFS server is
- shut down
- failed over
It may just be that we need a new way to fail them over, that might require unmounting everybody or using autofs.