Hi Dan,
I don't know why NBD wouldn't function, but I also don't think it's the
way you should go. Putting the journals on the OSD disks isn't a
terrible option, but you will suffer the expected double write penalty .
If your system disk is an SSD with fast sequential write throughput
capabilities, you might be able to put the journals there with minimal
overhead. If it's a spinning disk, I'd avoid putting the journals there
as you'll very quickly oversubscribe that disk.
Mark
On 10/31/2014 12:01 PM, Dan Ryder (daryder) wrote:
Hi cephers,
I’m designing a new “production-like” Ceph cluster, but I’ve run into an
issue. I have 4 nodes with 1 disk for OS, 3 disks for OSDs on each node.
However, I only have 2 extra disks for use of OSD journals.
My first question is if it is possible to use a remote disk partition
(currently using NBD) for OSD journals on nodes that don’t have access
to a local disk for journaling. I’ve run into some problems getting Ceph
to recognize the NBD disks as block-devices during ceph-deploy osd
activate. If it is possible, would this be a good design? I think we
might run into some performance issues using a network journal for some
OSDs.
If it isn’t recommended, should I use partitions of my OS disk on 2 of
my nodes for a journal until I can get some more disks for journals? I
imagine this would be better than storing the journal on the same disk
as an OSD.
Thanks,
Dan
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
