On Fri, Jan 3, 2014 at 2:02 AM, Andrei Mikhailovsky <andrei@xxxxxxxxxx> wrote: > Hi guys, > > Could someone explain what's the new perf stats show and if the numbers are > reasonable on my cluster? > > I am concerned about the high fs_commit_latency, which seems to be above > 150ms for all osds. I've tried to find the documentation on what this > command actually shows, but couldn't find anything. > > I am using 3TB sas drives with 4 osd journals on each ssd. Are the numbers > below reasonable for a fairly idle ceph cluster (osd utilisation below 10% > on average)? This is a report about flushing data out to the backing-store disk, and fs_commit_latency is generally going to include a syncfs syscall, so 150-600 ms is not unreasonable. The fs_apply_latency (for applying updates to the in-memory filesystem) and the numbers on the journal are the ones to look at. -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com > > # ceph osd perf > osdid fs_commit_latency(ms) fs_apply_latency(ms) > 0 192 4 > 1 265 4 > 2 116 1 > 3 125 2 > 4 166 1 > 5 209 3 > 6 184 6 > 7 142 2 > 8 209 1 > 9 166 1 > 10 216 1 > 11 308 3 > 12 150 2 > 13 125 1 > 14 175 2 > 15 142 2 > 16 150 4 > > > when the cluster get's a bit busy (osd utilisation below 50% on average) I > see: > > # ceph osd perf > osdid fs_commit_latency(ms) fs_apply_latency(ms) > 0 551 11 > 1 284 25 > 2 517 41 > 3 492 14 > 4 625 13 > 5 309 26 > 6 650 9 > 7 517 21 > 8 634 25 > 9 784 32 > 10 392 7 > 11 501 8 > 12 602 12 > 13 467 14 > 14 476 36 > 15 451 11 > 16 383 21 > > > Thanks > > Andrei > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |