The only test I’ve performed on the journal was running dstat on the journal and OSDs. I know the sustained throughput performance for the journal drive and OSD drives, then I keep writing small files (1MB) to Ceph cluster mounting it using Ceph kernel block device driver. If the journal throughput measured by dstat is close to the theoretical throughput but the OSDs are not, that means that the journal is throttling the cluster performance, and if the journal and OSDs are pushed to their limits, that means that the journal is doing OK. Hope that helps! Regards, Bishoy
|
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com