On 30.10.2012 14:47, Maciej Gałkiewicz wrote:
ServerA(mon+osd):
Filesystem Size Used Avail Use% Mounted on
/dev/sda1 9.2G 2.4G 6.4G 27% /
tmpfs 5.9G 0 5.9G 0% /lib/init/rw
udev 5.9G 148K 5.9G 1% /dev
tmpfs 5.9G 0 5.9G 0% /dev/shm
/dev/sda7 674G 483M 629G 1% /var/lib/ceph/mon
/dev/sda6 917G 2.4G 868G 1% /var/lib/ceph/osd/ceph-4
I see that you have separate partitions for osd and mon. Some of them
are on the same disk as OS. I would strongly recommend having separate
disk (not just partition) for each osd.
There is another problem probably more important. By default journal
is stored in /var/lib/ceph/osd/$cluster-$id/journal which means you
have data and journal on the same disk. It is what slows you down the
most. I was amazed how faster the cluster can be just by putting the
journal somewhere else (tmpfs - only for tests, another disk). From my
experiance, there is no need for separate disk for mon. It can go with
OS.
Run rados bench everytime you make an improvement in your cluster.
Oops, it's actually true!!!
Give me a moment to check this out and thank you for opening my eyes.
Let me to notify you about the results of implementing your recommendations.
I'm going to completely rebuild my cluster. Thank so much.
--
Kind regards,
R. Alekseev
--
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
