[Gluster-devel] Gluster 3.5 (latest nightly) NFS memleak

[Vijay Bellur]

On 03/27/2014 03:29 AM, Giuseppe Ragusa wrote:
> Hi all,
> I'm running glusterfs-3.5.20140324.4465475-1.autobuild (from published
> nightly rpm packages) on CentOS 6.5 as storage solution for oVirt 3.4.0
> (latest snapshot too) on 2 physical nodes (12 GiB RAM) with
> self-hosted-engine.
>
> I suppose this should be a good "selling point" for Gluster/oVirt and I
> have solved almost all my oVirt problems but one remains:
> Gluster-provided NFS (as a storage domain for oVirt self-hosted-engine)
> grows (from reboot) to about 8 GiB RAM usage (I even had it die before,
> when put under cgroup memory restrictions) in about one day of no actual
> usage (only the oVirt Engine VM is running on one node with no other
> operations done on it or the whole cluster).
>
> I have seen similar reports on users and devel mailing lists and I'm
> wondering how I can help in diagnosing this and/or if it would be better
> to rely on latest 3.4.x Gluster (but it seems that the stable line has
> had its share of memleaks too...).
>

Can you please check if turning off drc through:

volume set <volname> nfs.drc off

helps?

-Vijay