Support #41932
closed
nautilus: pool does not respond to pg_num setting
Added by George Kyriazis over 4 years ago.
Updated over 4 years ago.
Description
Installed mimic and upgraded every node to nautilus.
Trying to set pg_num and pgp_num to a higher setting, and setting is ignored. Also tried using autoscaler, it looks like autoscaler makes a decision to change pg_num, but no actual change happening.
/var/log/ceph# ceph osd pool set rbd1 pg_num 128
set pool 10 pg_num to 128
/var/log/ceph# ceph osd pool get rbd1 pg_num
pg_num: 100
/var/log/ceph#
Also
/var/log/ceph# ceph osd pool ls detail | grep rbd1
pool 10 'rbd1' replicated size 1 min_size 1 crush_rule 0 object_hash rjenkins pg_num 100 pgp_num 100 pg_num_target 128 pgp_num_target 128 autoscale_mode on last_change 3018 flags hashpspool,nearfull,selfmanaged_snaps,creating stripe_width 0 application rbd
/var/log/ceph#
(node discrepancy between pg_num and pg_num_target).
- Tracker changed from Bug to Support
Sounds like you missed an upgrade step that enabled the full Nautilus features so that the cluster could start doing splits and merges based on the PG target?
It's only one pool (the rbd pool) that has (had) this issue. Other pools can resize with no problem.
I ended up recreating the pool, and the new pool now works fine.
So, what would the correct upgrade process, then? I upgraded using apt, first the OSD nodes, then the monitors/managers and last the admin node. Reboot after installing on every node.
- Status changed from New to Closed
yes, I did set require-old-release to nautilus, but still.
In any case, since I recreated the pool, I don't have this issue anymore, but I was wondering if you guys wanted to investigate further. Looks like that's not the case, since you closed it. :-)
Also available in: Atom
PDF