180PG per OSD is usually overhead, also 40k obj per PG is not much, but I don't think this will works without block.db NVMe. I think your "wrong out marks" evulate in time of rocksdb compaction. With default log settings you can try to grep 'latency' strings
Also, https://tracker.ceph.com/issues/50297 k Sent from my iPhone > On 22 Mar 2022, at 14:29, Boris Behrens <[email protected]> wrote: > * the 8TB disks hold around 80-90 PGs (16TB around 160-180) > * per PG we've around 40k objects 170m objects in 1.2PiB of storage _______________________________________________ ceph-users mailing list -- [email protected] To unsubscribe send an email to [email protected]
