Gluster users - Mar 2019 - Possible memory leak via wordpress wordfence plugin behavior in 4.1.16

I have 4 machines running glusterfs and wordpress with the wordfence plugin.

The wordfence plugin in all 4 instance pounds away writing and
re-writing the file:

/mnt/glusterfs/www/openvpn.net/wp-content/wflogs/config-synced.php

This is leading to stale file handle errors reported by glusterfs and
while the request is ultimately handled correctly this looks to be
leading to a memory leak.

I think the leak is dict_t structures based on looking at dump state
but my knowledge of glusterfs is mostly as a user.

--
brian

On 09/03/19 7:15 AM, Brian Litzinger wrote:
> I have 4 machines running glusterfs and wordpress with the wordfence
plugin.
>
> The wordfence plugin in all 4 instance pounds away writing and
> re-writing the file:
>
> /mnt/glusterfs/www/openvpn.net/wp-content/wflogs/config-synced.php
>
> This is leading to stale file handle errors reported by glusterfs and
> while the request is ultimately handled correctly this looks to be
> leading to a memory leak.
Memory leak of the glusterfs client process?
What type of volume (`gluster volume info`) are you using and how many 
clients are accessing the volume? Can you share the client(s) and the 
bricks logs when you see the ESTALE error? Also share successive state 
dump outputs of the process where you suspect dict_t leaks.

Thanks,
Ravi

> I think the leak is dict_t structures based on looking at dump state
> but my knowledge of glusterfs is mostly as a user.
>
> --
> brian
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> https://lists.gluster.org/mailman/listinfo/gluster-users