[Gluster-devel] possible memory leak when 'kill -HUP glusterfs'

[Kenta Takahashi]

Hi,

We've found possible memory leak when kill -HUP the glusterfs process.

ENVIRONMENT:
OS: RHEL6 x86_64
GlusterFS: 3.2.5

HOW TO REPRODUCE:
1. install glusterfs
2. start glusterd
3. create and start a volume
4. repeat 'killall -HUP glusterfs'

Then you can see the glusterfs process keeps increasing its memory space until entire RAM/Swap space exhausted,
and die (killed?) silently (no logs generated).

We are using 'kill -HUP glusterfs' processes when logrotate done like below:
https://github.com/gluster/glusterfs/blob/master/extras/glusterfs-logrotate

Any ideas to workaround/fix this ?

-- 
Kenta Takahashi
knt.takahashi at ntt.com