Hi, we recently upgraded one of our Ceph clusters from Infernalis to Jewel. The upgrade process went smoothly: Upgraded OSD's, restarted them in batches, waited for health OK, updated Mon and MDS, restarted, waited for heatlh OK, etc. We then set require_jewel_osds flag and upgraded our CephFS clients. It seemed like the actual upgrade process went really well.
Except for the fact that our write performance for CephFS clients has dropped off a cliff. Previously, we were writing at 1-1.5 GB/sec (bytes not bits). After upgrading to Jewel, it's now in the 200-300 MB/sec range. Are there any known issues with Jewel sequential write performance? Is there some sort of upgrade task running that's limiting throughput?
There are no errors in the logs, and I'm really struggling to find information on what might have changed other than the Ceph version.
TIA,
-richard
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
