Hello, On Mon, 10 Dec 2018 20:43:40 -0500 Tyler Bishop wrote: > I don't think thats my issue here because I don't see any IO to justify the > latency. Unless the IO is minimal and its ceph issuing a bunch of discards > to the ssd and its causing it to slow down while doing that. > What does atop have to say? Discards/Trims are usually visible in it, this is during a fstrim of a RAID1 / : --- DSK | sdb | busy 81% | read 0 | write 8587 | MBw/s 2323.4 | avio 0.47 ms | DSK | sda | busy 70% | read 2 | write 8587 | MBw/s 2323.4 | avio 0.41 ms | --- The numbers tend to be a lot higher than what the actual interface is capable of, clearly the SSD is reporting its internal activity. In any case, it should give a good insight of what is going on activity wise. Also for posterity and curiosity, what kind of SSDs? Christian > Log isn't showing anything useful and I have most debugging disabled. > > > > On Mon, Dec 10, 2018 at 7:43 PM Mark Nelson <mnelson@xxxxxxxxxx> wrote: > > > Hi Tyler, > > > > I think we had a user a while back that reported they had background > > deletion work going on after upgrading their OSDs from filestore to > > bluestore due to PGs having been moved around. Is it possible that your > > cluster is doing a bunch of work (deletion or otherwise) beyond the > > regular client load? I don't remember how to check for this off the top > > of my head, but it might be something to investigate. If that's what it > > is, we just recently added the ability to throttle background deletes: > > > > https://github.com/ceph/ceph/pull/24749 > > > > > > If the logs/admin socket don't tell you anything, you could also try > > using our wallclock profiler to see what the OSD is spending it's time > > doing: > > > > https://github.com/markhpc/gdbpmp/ > > > > > > ./gdbpmp -t 1000 -p`pidof ceph-osd` -o foo.gdbpmp > > > > ./gdbpmp -i foo.gdbpmp -t 1 > > > > > > Mark > > > > On 12/10/18 6:09 PM, Tyler Bishop wrote: > > > Hi, > > > > > > I have an SSD only cluster that I recently converted from filestore to > > > bluestore and performance has totally tanked. It was fairly decent > > > before, only having a little additional latency than expected. Now > > > since converting to bluestore the latency is extremely high, SECONDS. > > > I am trying to determine if it an issue with the SSD's or Bluestore > > > treating them differently than filestore... potential garbage > > > collection? 24+ hrs ??? > > > > > > I am now seeing constant 100% IO utilization on ALL of the devices and > > > performance is terrible! > > > > > > IOSTAT > > > > > > avg-cpu: %user %nice %system %iowait %steal %idle > > > 1.37 0.00 0.34 18.59 0.00 79.70 > > > > > > Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s > > > avgrq-sz avgqu-sz await r_await w_await svctm %util > > > sda 0.00 0.00 0.00 9.50 0.00 64.00 > > > 13.47 0.01 1.16 0.00 1.16 1.11 1.05 > > > sdb 0.00 96.50 4.50 46.50 34.00 11776.00 > > > 463.14 132.68 1174.84 782.67 1212.80 19.61 100.00 > > > dm-0 0.00 0.00 5.50 128.00 44.00 8162.00 > > > 122.94 507.84 1704.93 674.09 1749.23 7.49 100.00 > > > > > > avg-cpu: %user %nice %system %iowait %steal %idle > > > 0.85 0.00 0.30 23.37 0.00 75.48 > > > > > > Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s > > > avgrq-sz avgqu-sz await r_await w_await svctm %util > > > sda 0.00 0.00 0.00 3.00 0.00 17.00 > > > 11.33 0.01 2.17 0.00 2.17 2.17 0.65 > > > sdb 0.00 24.50 9.50 40.50 74.00 10000.00 > > > 402.96 83.44 2048.67 1086.11 2274.46 20.00 100.00 > > > dm-0 0.00 0.00 10.00 33.50 78.00 2120.00 > > > 101.06 287.63 8590.47 1530.40 10697.96 22.99 100.00 > > > > > > avg-cpu: %user %nice %system %iowait %steal %idle > > > 0.81 0.00 0.30 11.40 0.00 87.48 > > > > > > Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s > > > avgrq-sz avgqu-sz await r_await w_await svctm %util > > > sda 0.00 0.00 0.00 6.00 0.00 40.25 > > > 13.42 0.01 1.33 0.00 1.33 1.25 0.75 > > > sdb 0.00 314.50 15.50 72.00 122.00 17264.00 > > > 397.39 61.21 1013.30 740.00 1072.13 11.41 99.85 > > > dm-0 0.00 0.00 10.00 427.00 78.00 27728.00 > > > 127.26 224.12 712.01 1147.00 701.82 2.28 99.85 > > > > > > avg-cpu: %user %nice %system %iowait %steal %idle > > > 1.22 0.00 0.29 4.01 0.00 94.47 > > > > > > Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s > > > avgrq-sz avgqu-sz await r_await w_await svctm %util > > > sda 0.00 0.00 0.00 3.50 0.00 17.00 > > > 9.71 0.00 1.29 0.00 1.29 1.14 0.40 > > > sdb 0.00 0.00 1.00 39.50 8.00 10112.00 > > > 499.75 78.19 1711.83 1294.50 1722.39 24.69 100.00 > > > > > > > > > _______________________________________________ > > > 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 > > -- Christian Balzer Network/Systems Engineer chibi@xxxxxxx Rakuten Communications _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com