On Tue, Feb 26, 2013 at 4:35 PM, <ms@xxxxxxxxxxxx> wrote: > Hello, > > I was wondering whether it would be feasible to manage existing FC-SAN > storage with ceph. Now this may sound somewhat weird, so let me explain: > > as it turns out you can't actually trust SAN boxes with RAID-6 devices > to actually hold your data, so I thought that maybe you could > > - set the existing SAN boxes to export its storage in JBOD chunks > - use some blades (they have 64gigs of RAM each) to mount those SAN FC > exports > - use those mounts as OSDs > - and store the data on 3 OSDs (each one being a different SAN box) to > get HA back > > Or maybe it would be better to set up the SAN boxes to export RAID-5 > devices and only have the data reside on 2 OSDs then? Both will work. Your choice might come down to performance of your SANs in JBOD vs. RAID-5. If you can assign any SAN cache (assuming its battery backed) to a particular LUN, you can make a small(ish) device and put your osd journal there, which might really boost your performance. > > Is there anything particularly problematic with this idea? We've seen other users run with similar setups successfully. > > Since the SANs actually have SAS disks, I thought that one might get > away with having no SSD in the blade for an intent log...? I think it depends on your performance/cost target. An SSD in the blade can still be faster than roundtrips over fc to your SAN, esp. with some of the PCIe SSD cards out there now. You have a variety of choices, but a good starting point is measuring your existing SAN performance and going from there. -slang [inktank developer] > > > > Regards, ms > _______________________________________________ > 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
 |