Yep, it happen on every run, I have checked on other VMs that do not use Ceph, it had almost no delay, although results were about similarly, 576 iops for Ceph's VM and 650 for non-Ceph VM, I use one image for all test, ubuntu 14.04.1, kernel 3.13.0-32-generic
2015-12-31 23:51 GMT+07:00 Jan Schermer <[email protected]>: > Is it only on the first run or on every run? > Fio first creates the file and that can take a while depeding on how > fallocate() works on your system. In other words you are probably waiting > for a 1G file to be written before the test actually starts. > > Jan > > > On 31 Dec 2015, at 04:49, Sam Huracan <[email protected]> wrote: > > Hi Ceph-users. > > I have an issue with my new Ceph Storage, which is backend for OpenStack > (Cinder, Glance, Nova). > When I test random write in VMs with fio, there is a long delay (60s) > before fio begin running. > Here is my script test: > > fio --directory=/root/ --direct=1 --rw=randwrite --bs=4k --size=1G > --numjobs=3 --iodepth=4 --runtime=60 --group_reporting --name=testIOPslan1 > --output=testwriterandom > > It is so strange, because I have not ever seen this method when I test in > Physical machine. > > My Ceph system include 5 node, 2 SAS 15k per node, I use both journal > and Filestore in one disk, Public Network 1 Gbps/Node, Replica Network 2 > Gbps/Node > > Here is my ceph.conf in node compute: > http://pastebin.com/raw/wsyDHiRw > > > Could you help me solve this issue? > > Thanks and regards. > > > > _______________________________________________ > ceph-users mailing list > [email protected] > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > >
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
