Hi All, Thought I would just share this as it's taken me a couple of hours to work it out. I'm using Diamond+Carbon+Graphite to collect the perf dump output from the admin sockets. The Diamond collector is really good at capturing everything, but stores the sum+avgcount directly into Carbon. In Graphite 0.10.0 there are new map+reduce functions which will finally let you calculate the average (or min/max.etc) latency over all of your OSD's with just one line. Ignore any info suggesting it's in the 0.9.x series.it's not :-( I'm running Graphite 0.9.15 as its packaged, but it was fairly easy just to copy the new functions into the python source, nothing else needed changing. Anyway here is the one liner scale(averageSeries(reduceSeries(mapSeries(aliasByNode(nonNegativeDerivative (servers.*.CephCollector.ceph.osd.*.osd.op_w_latency.*),1,2,3,4,5,6,7,8),1), "divideSeries",7,"sum","avgcount")),1000) This will work for any of the counters that have the sum+avgcount format Hope that helps someone Nick Ps. If anyone has a link, describing what each of the ceph perf counters do in more detail, I would be very happy if you could share it. Otherwise I shall probably ask a lot of questions and then write something up myself. _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |