Hello everybody,
we have a number of "replica 3 arbiter 1" or (2 + 1) volumes
because we're running out of space on some volumes I need to optimize
the usage of the physical disks. that means I want to consolidate
volumes with low usage onto the same physical disk. I can do it with
"replace-brick commit force" but that looks a bit drastic to me
because it immediately drops the current brick and rebuilds the new
one from the remaining bricks. Is there a possibility which builds the
new brick in the background and changes config only when it's fully in
sync?
I was thinking about
- dropping arbiter brick => replica 2
- adding a new brick => replica 3
- dropping old brick => replica 2
- re-adding arbiter brick => replica 2 arbiter 1
About 20 years ago, I was managing Vertitas Volume Manager. To move a
sub-disk (= similar to brick) VVM temporarily upgraded the subdisk to
a mirrored volume, synced both sides of the mirror and then downgraded
the construct to the new sub-disk. it was impressive and scary at the
same time but we never had an outage.
BTW: I'm running Gluster 3.8.15
BTW: new storage is ordered but the reseller fucked up and now we have
to wait for the delivery for 2 months
Kind Regards
Bernhard
