James Harper , 6/21/2013 7:17 AM: > Hi all > > I'm building a small ceph cluster with 3 nodes (my first ceph cluster). > Each Node with one System Disk, one Journal SSD Disk and one SATA OSD > Disk. > > My question is now should I use Desktop or Enterprise SATA Drives? > Enterprise Drives have a higher MTBF but the Firmware is actually build for > RAID Controllers > which have different timing. WD calls that TELR, Seagate Error Recovery > Control (ERC). > In case of an error, they make shure the RAID Controller doesn't throw them > out of the array after short time.
>>Do you mean that the RAID controller _does_ through them out after a very >>short time? No, not exactly. After a very short time the Drive signals the RAID Controller that it has a problem and need more time to fix it. The Controller read it the from another drive. In the meantime the Drive can fix it without get trought out from the RAID. At least is this what I understand from the Infos on the Vendors side. > Desktop Drives can try seconds to minutes until they give up. > Can Rados/Ceph handle such long time without get answer from the Drive? > Will the Node then marked bad/out from the cluster? > >>Your users might not like the delays. For a single disk it makes sense to >>retry for a while to try and give the user their data. For RAID it makes no >>sense. Fail immediately and eject the disk and don't make delays. >>I wonder if those settings can be tweaked at all? I think smartctl can do that. But not 100% sure. Earlier WD had a tool to do that but they removed it from the download site. >>James Stefan
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
