Am Wed, 30 Jun 2010 14:05:34 +0200 schrieb Sébastien Paolacci: > Hello Sage, > > It actually works better when removing the old deb pacakge;)... thanks! > > The "pg 0.xxxx on [] is laggy" messages have also now disappeared. > > Here is, just for information, the obtained two processes dbench output > (which is not so bad given the vm the system in running into): Operation > Count AvgLat MaxLat > ---------------------------------------- NTCreateX 1632 8.835 > 213.966 Close 1237 0.377 157.360 Rename 73 > 24.290 195.969 Unlink 291 6.514 244.166 Qpathinfo > 1455 4.544 237.949 Qfileinfo 311 0.233 5.723 > Qfsinfo 270 1.854 110.355 Sfileinfo 168 3.090 > 34.291 Find 552 11.182 192.610 WriteX 1076 > 4.916 409.798 ReadX 2437 0.408 61.635 LockX > 4 0.011 0.017 UnlockX 4 0.006 0.009 > Flush 140 8.630 175.423 > > Throughput 2.88849 MB/sec 2 clients 2 procs max_latency=409.805 ms > > I however get the a bunch of metadata sync warnings on the cfuse side > "10.06.30_13:16:20.556187 7f2bc5817720 client4100 fsync - not syncing > metadata yet.. implement me" give the kclient a try? git://ceph.newdream.net/ceph-client- standalone.git branch unstable-backport for use with "older" kernels. > [osd] > debug ms = 1 > sudo = true > osd data = /data/osd$id > osd journal = /ceph-journal/osd$id why not let ceph manage the whole partition without a fs inbetween ? like osd journal = /dev/sdX1 (and remove "journal size") > filestore journal writeahead = true seems to be true by default (http://ceph.newdream.net/wiki/Ceph.conf -> filestore_journal_parallel) - Thomas -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
