Bug 1133073

Description of problem:
I have a cluster with 2 node replicas. 
In cluster we have 1 brick\volume. Capacity 44Tb. Used 27Tb.
If some times do recursive listing of files (ls -laR) on volume, processes glusterfs get memory and don't return it. After some cycle recursive listing all memory used by glusterfs. It can be more 50Gb on server with 64Gb RAM. After that occurs OOM.

Try dissable options:
performance.quick-read
performance.io-cache
but it didn't give good result.

echo 2 > /proc/sys/vm/drop_caches 
didn't help in this situation.

As temporary solution use umount and mount volume when memory exhausted.


Version-Release number of selected component (if applicable):
OS: Ubuntu 12.04.4 LTS \n \l
Core: 3.10.28-6
Packages:
ii  glusterfs-client                     3.4.4-5                           clustered file-system (client package)
ii  glusterfs-common                     3.4.4-5                           GlusterFS common libraries and translator modules
ii  glusterfs-server                     3.4.4-5                           clustered file-system (server package)


How reproducible:
Some times do recursive listing on files in volume.

Steps to Reproduce:
1. Create replica set from 2 nodes with 1 brick. 
2. Add files and get big usage capacity volume. We have total copacity 44T and used 27T.
3. Do some times recursive listing on all mounted volume.

Actual results:
Exhaustion memory (RAM)

Expected results:


Additional info:
# gluster volume info
 
Volume Name: repofiles
Type: Replicate
Volume ID: 34f85192-2b9a-4d36-8468-3576d0cc922a
Status: Started
Number of Bricks: 1 x 2 = 2
Transport-type: tcp
Bricks:
Brick1: xxx1:/mnt/gluster_source/brick
Brick2: xxx2:/mnt/gluster_source/brick
Options Reconfigured:
performance.read-ahead: on
performance.write-behind-window-size: 4MB
nfs.disable: on
performance.nfs.stat-prefetch: off
performance.nfs.io-threads: off
performance.nfs.read-ahead: on
performance.nfs.io-cache: on
performance.stat-prefetch: on
performance.client-io-threads: off
performance.io-thread-count: 32
performance.quick-read: off
performance.io-cache: off
performance.cache-size: 6442450944
performance.nfs.quick-read: on