[Bugs] [Bug 1403984] Node node high CPU - healing entries increasing

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1403984

Ravishankar N <ravishankar at redhat.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tu2Bgone at gmail.com
              Flags|                            |needinfo?(tu2Bgone at gmail.co
                   |                            |m)



--- Comment #2 from Ravishankar N <ravishankar at redhat.com> ---
High CPU usage is a known problem, we are working on introducing some form of
throttling mechanism to regulate self-heal traffic.

Meanwhile, some queries:
1. What process was consuming high CPU (was it brick/glustershd/glusternfs
process)?
2. How many files were pending heal when you noticed the high cpu load (i.e.
'Number of entries' in each brick when `gluster volume heal volname info` is
run)
3. The [Operation not permitted] should normally not occur. Can you share the
complete brick log? Also, the filename corresponding to the gfid
9b886490-6a45-433f-aa49-a398c4ef385d. Is there a difference in file
owner/permissions/times for this file amongst the 3 bricks? 


Also, some tips:
1. cluster.quorum-type: none --> This has to be set to 'auto' to avoid
split-brains.
2. Can you try using cgroups for regulation and let us know if that helps? (
http://blog.gluster.org/2016/09/making-gluster-play-nicely-with-others/ )

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=H823d5qAwD&a=cc_unsubscribe