Hi everyone:
I use ceph-deploy to configure SSD Journal.
ceph-deploy osd create ceph-node:sdb1:/dev/ssd1 ceph-node:sdc1:/dev/ssd2
At the first test, each osd have a 5G ssd journal partition.
I used fio to test performance:
fio -filename=/dev/rbd0 -direct=1 -iodepth 32 -thread -rw=randwrite -ioengine=libaio -bs=64k -size=2000G -group_reporting -name=mytest
the performance of rados is instability. at the beginning almost 700MB/s, but drop to 50MB/s periodicity.
I guess maybe 5G is too small for SSD journal. So I create 100G SSD journal partition for each OSD.
I test again, and find a very good result, the problem solved.
So i want to know what is the suitable size for SSD Journal.
I read the document about Journal configure.
osd journal size = {2 * (expected throughput * filestore max sync interval)}the default value for filestore max sync interval is 5 second.I didn't understand "expected throughput" mean? It is mean a osd disk max throughput?
If this, I think 5G is enough.Maybe "expected throughput" mean a max throughput for SSD disk?
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
