Hi guys, I'm invastigating rados object latency ( with /usr/bin/time -f"%e" rados -p chunks get $chk /dev/shm/test/test.file ). Objects are around 7MB +-1MB on size. Results shows that 0.50% of object are fetched from cluster in 1-4 seconds and the rest of objects are good, below 1sec (test is done on 5000 objects and results are replicable). Most of these slow objects according osd map belongs to 3rd and 4th osd node. So I'm trying to find what is wrong with these two servers or what could/should be tuned. I measured lookup and 0.03% are slow (1-4s) which is acceptable for now. Cluster contains 5 mons a 4 osd nodes 2 nodes x30 osds, 2 x 20 osds and same 3TB WD discs 4098pgs. Is this latency normal? Where could be the bottleneck? Thanks
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
