[Gluster-users] High CPU Usage After Glusterfs install

[Brian Candler]

On Wed, Mar 14, 2012 at 03:13:29PM -0400, Joseph Hardeman wrote:
>    Thank you for responding.  So you aren't using the vol files in
>    /etc/glusterfs to control anything, such as afra or unity?

Nope - indeed I have no idea what afra or unity are (and googling for
"gluster afra unity" doesn't match anything useful)

I have used the CLI utils as per the documentation, and everything "just
works".  There are a three files in /etc/glusterfs/ but they have not
changed:

$ ls -l /etc/glusterfs/
total 12
-rw-r--r-- 1 root root  229 2011-11-18 07:00 glusterd.vol
-rw-r--r-- 1 root root 1908 2011-11-18 07:00 glusterfsd.vol.sample
-rw-r--r-- 1 root root 2005 2011-11-18 07:00 glusterfs.vol.sample

All the config changes are instead reflected under /etc/glusterd/

$ ls /etc/glusterd/
geo-replication  glusterd.info  nfs  peers  vols

I see there's lots of *old* documentation for gluster <=2.x which talks
about doing things manually, but the new documentation has been seriously
dumbed down and doesn't even mention the module stacking configuration
files.

>    I am just
>    asking because after building my own rpms and installing them, I was
>    able to build like I did before and I didn't see the high CPU usage.
>    Now the weird thing I saw was during a test failover and
>    stopping/starting glusterd on the first of pair I did see high cpu and
>    the vm's hung.

Attaching strace to the gluster processes might give you an idea what's
happening?

Regards,

Brian.