--- Begin Message ---
Hi,
We upgraded a cluster to Reef yesterday and we’re suffering from laggy
OSD’s, a lot. There seems to be an issue where RocksDB is looking for
64k blocks that are aligned, but cannot find them (fast enough). The OSD
starts logging stuff like:
2024-01-12T10:01:59.513+0100 7f339ae146c0 0
bluestore(/var/lib/ceph/osd/ceph-3) log_latency_fn slow operation
observed for _txc_committed_kv, latency = 37.572166443s, txc =
0x5609c5edcc00
This issue seems to exist for a while already, but did not show up,
which is ‘fixed’ in Reef. Someone close to the Ceph community advises me
to upgrade to 18.2.1, but it’s not in Proxmox Enterprise yet.
What’s the ETA on that? I would like to keep the cluster as ‘enterprise’
as possible and refrain from mixing repo’s.
Relevant Ceph tracker links, which I understand little of because I’m
not smart enough to be a developer of Ceph:
https://tracker.ceph.com/issues/63760
https://tracker.ceph.com/issues/62815
Thanks!
—
Mark Schouten
CTO, Tuxis B.V.
+31 318 200208 / m...@tuxis.nl
--- End Message ---
_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
