On 1/10/19 12:59 PM, Marc wrote: > Hi, > > for support reasons we're still running firefly (part of MCP 6). In our > grafana monitoring we noticed that two out of 128 OSD processes show > significantly higher outbound IO than all the others and this is > constant (cant see first occurance of this anymore, grafana only has 14 > days backlog). Those 2 show 350MB/s out whereas all the other are at > 50MB/s - constantly. There's obviously spikes that go above that but > that seems to be idle / normal rate throughout the day. > > Is there a good way to debug this with standard tools (i.e. no sysdig / > nmon)? See which IP this goes to and/or what object/s this is caused by? > Ceph is only serving KVM images as is default for most installations I > suppose. > You could increase debug ms and debug rados on the OSDs and see which operations they are doing. Try with 10 and 20 for both values to see what the OSDs are doing. Wido > > Thanks! > M. > _______________________________________________ > 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
 |