Gluster users - Aug 2016 - 3.8.2

Upgraded from 3.8.1 to 3.8.2 over the weekend - no issues so far.

Have set op-version 30800
cluster.granular-entry-heal = on
cluster.locking-scheme = granular

I initially did a rolling ugrade, updating each server, waiting to
heal, then the next. Was quite straightforward.

Then I did a clean boot of all servers - because I can.


-- 
Lindsay

On 8/14/16 20:28, Lindsay Mathieson wrote:
> Then I did a clean boot of all servers - because I can.
>
Heheh, spit my soda out on that one.

-wk

On 15 August 2016 at 13:28, Lindsay Mathieson
<lindsay.mathieson at gmail.com> wrote:
> Upgraded from 3.8.1 to 3.8.2 over the weekend - no issues so far.

Ok, as an experiment I killed glusterfsd on one node. System kept
chugging allong fine with no hiccups. I ran a few disk intensive VM's
on that node and others, no real slow down. Monitoring it with "heal
statistics heal-count'

heal-count got up to approx 2500 shards and restarted glusterfsd by
restarting the gluster-service (glusterd).

heal-count stopped rising, but what is concerning is that it doesn't
seem to be going back down. 45min later at its stable at 2439 files
needing healed and glusterfsd is thrashing the CPU's on that node
(1000%!)

The glfsheal log has no entries at all.

Previously (3.7.x) when I've done this test, heals kicked in very rapidly.

-- 
Lindsay