Don’t use legacy override reweights. When you have upmap balancing enabled they confuse the balancer. Look for `rados bench` leftovers
On May 5, 2025, at 6:01 AM, Yunus Emre Sarıpınar <yunusemresaripinar@gmail.com> wrote:
I have a ceph cluster created with nautilus and I upgraded to octopus later.
I was have 24x node and I added 8x new nodes to the cluster. Balancer is upmap enabled. I increased my PG count from 8192 to 16384.
I had to set reweight 0.8 on new OSD's to solve full usage (There was only 1TB left because of the faulted balance)
Now cluster health is OK and distribution is balanced by my manual fix.
data: pools: 4 pools, 16545 pgs objects: 173.32M objects, 82 TiB usage: 436 TiB used, 235 TiB / 671 TiB avail pgs: 15930 active+clean 579 active+clean+snaptrim_wait 36 active+clean+snaptrim
Before adding 8 nodes to the cluster, the usage was 215 TiB used. I did the update about 2 months ago and my usage area still hasn't decreased.
Why did the usage double, how can I solve it?
_______________________________________________ Dev mailing list -- dev@ceph.io To unsubscribe send an email to dev-leave@ceph.io