[Gluster-users] glusterfs and glusterfsd process utilization

[Kyle Harris]

Hello All,

I have been experiencing difficulty with high process utilization on the
glusterfs and glusterfsd processes.  I thought I had it tracked down to
Apache looking for files on the cluster drive that don't exist.  I have
been working to clean that up and now I have discovered another problem.

When I reboot one of the nodes in my 3-node replicated cluster, as soon as
it comes back up I see a bunch of entries such as follows:

[2014-12-21 05:33:49.873989] I [dict.c:370:dict_get] (-->
/usr/lib64/libglusterfs.so.0(_gf_log_callingfn+0x1e0)[0x7fe3bd0d2420] (-->
/usr/lib64/libglusterfs.so.0(dict_get+0x4f)[0x7fe3bd0cbbdf] (-->
/usr/lib64/glusterfs/3.6.1/xlator/system/posix-acl.so(posix_acl_lookup_cbk+0x1e1)[0x7fe3aad3ada1]
(-->
/usr/lib64/glusterfs/3.6.1/xlator/debug/io-stats.so(io_stats_lookup_cbk+0x113)[0x7fe3aaf524e3]
(-->
/usr/lib64/glusterfs/3.6.1/xlator/performance/md-cache.so(mdc_lookup+0x300)[0x7fe3ab167310]
))))) 0-dict: !this || key=system.posix_acl_access

While this is occurring, the process utilization for glusterfs and
glusterfsd go sky high.  Eventually they calm down on their own but the
servers are virtually unusable until they do.  Note this is on Centos 6.6
with Gluster 3.6.1 and kernel 2.6.32 using a fuse mount with each of the 3
servers having the client mounted for hosting web sites via Apache.  There
are many small files and some are php.  I have not done any tuning to the
default gluster parameters.

Can someone tell me why this occurs after one of the nodes is rebooted and
perhaps lend some guidance on what can be done to resolve this issue?

Thank you.

-- 
Kyle
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20141220/dedb19cc/attachment.html>