[Gluster-users] gluster heal performance
mbaehr+gluster at realss.com
Fri Sep 11 09:12:50 UTC 2020
Excerpts from Gionatan Danti's message of 2020-09-11 08:34:04 +0200:
> > we have a 3-node replica cluster running version 3.12.9
> > with 32TiB of space. each node has a single brick on
> > top of a 7-disk raid5 (linux softraid)
> 3.12.9, while being the official RHEL 7 release, is very old now.
yes, i am aware. we didn't bother upgrading as we need to expand
capacity and it's cheaper to rent new servers than expand the old ones.
> > the accumulated changes of one month of operation took 4 months to
> > heal.
> Wow, 4 months is a lot... but you had at least internal redundancy
> (RAID5 bricks).
right, that, and we had 3 replicas. we could have just dropped the third
node, and would still have been ok.
for the new cluster we decided that 2 nodes is enough, because the data
is all backups anyways. even if we loose both nodes, we can at least in
theory still recover all the data. whether that's a good decision is a
risk calculation. is a third server worth the extra expense? we decided
that, for what is essentially a backup, it's not.
i considered 3 nodes but dropping the raid instead, but several comments
inclusing yours convinced me that keeping the raid is good. on the new
servers we'll each have 3 bricks with 5 disks in a raid 5 per brick.
> > the initial heal count was about 6million files for one node and
> > 5.4million for the other.
> > ...
> > we do have a few huge directories with 250000, 88000, 60000 and 29000
> > subdirectories each. in total 26TiB of small files, but no more than
> > a few 1000 per directory. (it's user data, some have more, some have
> > less)
> > could those huge directories be responsible for the slow healing?
> The very high number of to-be-healed files surely has a negative impact
> on your heal speed.
that sounds like that there is an inefficiency within the healing
process that causes the healing speed to be non-linear depending on the
number of files.
general manager realss.com
student mentor fossasia.org
community mentor blug.sh beijinglug.club
pike programmer pike.lysator.liu.se caudium.net societyserver.org
Martin Bähr working in china http://societyserver.org/mbaehr/
More information about the Gluster-users