Hi
We currently have a 3 DC setup with 6 HDD servers in each of 3 DCs,
running replica 3 and EC 4+5 pools. This works fine for mostly dormant
data but of course not so well for other stuff requiring low latency
etc. Mostly big RBDs shared by kernel NFS with a bit of CephFS.
We want to start migrating some of these big RBD backed NFS shares to
NVMe. We currently have 3 E3.S servers, one in each DC, with an EC 4+5
pool with DC->OSD crush selection. We are wondering how to cheapest
increase redundancy short term since we do not have enterprise money.
With what we have we can lose 1 DC (aka host in this case) and 1 OSD and
still be online, but there's nowhere to backfill and some hardware
problems, like a faulty mainboard or CPU can take down an entire "DC".
Ideally we would have several more hosts in each DC so we had
DC->Host->OSD but that's not an option.
I am thinking it could make sense to add just 1 more server in a 4th DC
and keep the 4+5 rule as is, simply giving us a buffer of 1 more failed
host before we hit problems. Thoughts?
Mvh.
Torkil
--
Torkil Svensgaard
Sysadmin
MR-Forskningssektionen, afs. 714
DRCMR, Danish Research Centre for Magnetic Resonance
Hvidovre Hospital
Kettegård Allé 30
DK-2650 Hvidovre
Denmark
Tel: +45 386 22828
E-mail: torkil@xxxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
