Hi,
The common recommendation is to use a good (Intel S3700) SSD disk for
journals for each 3-4 OSDs, or otherwise to use internal journal on each
OSD. Don't put more than one journal on the same spinning disk.
Also, it is recommended to use 500G-1TB disks, specially if you have a
1gbit network; otherwise when a OSD fails recover time can be quite
long. Also look in the mailing list archives for some tunning of
backfiling for smalls ceph clusters.
Cheers.
Eneko
On 06/02/15 16:48, pixelfairy wrote:
3 nodes, each with 2x1TB in a raid (for /) and 6x4TB for storage. all
of this will be used for block devices for kvm instances. typical
office stuff. databases, file servers, internal web servers, a couple
dozen thin clients. not using the object store or cephfs.
i was thinking about putting the journals on the root disk (this is
how my virtual cluster works, because in that version the osds are 4G
instead of 4TB), and keeping that on its current raid 1, for
resiliency but im worried about making a performance bottleneck.
tempted to swap these out with ssds. if so, how big should i get? is
1/2TB enough?
the other thought was little partitions on each osd. were doing xfs
because i dont know enough about brtfs to feel comfortable with that.
would the performance degredation be worse?
is there a better way?
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
--
Zuzendari Teknikoa / Director Técnico
Binovo IT Human Project, S.L.
Telf. 943575997
943493611
Astigarraga bidea 2, planta 6 dcha., ofi. 3-2; 20180 Oiartzun (Gipuzkoa)
www.binovo.es
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
