There were some allocator changes in bluestore afaik, but slower than
100 iops sounds pretty off unless you have a very small cluster with
HDDs. How are you benchmarking, and have you tried without ISCSI?
Mark
On 7/19/21 11:41 AM, Philip Brown wrote:
I had recently setup a test cluster of Ceph Octopus, on a particular set of hybrid OSD nodes.
It ran at a particular rated IO level, judging by "fio".
Now in the last month or so, I got to deploy an evaluation cluster of ceph pacific, on the same hardware.
It is *drastically* slower, using the same hardware. In some use cases, slower than 100io/s.
Could anyone suggest a reason for this? (and ideally, how to retune?)
Did the required minimum effective size of the WAL on SSD grow between releases, for example?
Target use is an iSCSI storage pool.
--
Philip Brown| Sr. Linux System Administrator | Medata, Inc.
5 Peters Canyon Rd Suite 250
Irvine CA 92606
Office 714.918.1310| Fax 714.918.1325
pbrown@xxxxxxxxxx| www.medata.com
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
