memory leak in glusterfs-client 3.13.2-1build1

[henk]

When I write from two glusterfs clients to the same file on the gluster server. The memory consumption on both clients increases until the client runs out of memory.
This bug was not present in glusterfs client 3.4.2-1ubuntu1. Also an upgrade to stock glusterfs 5.3-ubuntu1~bionic1 seems to fix this issue.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Dani García (dem.corp)]

Hi,

Same problem here using the same version in Ubuntu 18.04. While using rsync on big directories I get OOM error because of it.

[Juergen Stawska (justaw)]

Hey,

I'm also affected.
During extensive rclone sessions, the glusterfs has been killed by oom reaper occasionally.
Even an hourly "/bin/echo 2 > /proc/sys/vm/drop_caches" does not circumvent the crash reliably.

According to slaptop "fuse_inode" consumes a growing amount of slab cache.
---------------------------
storage side details:

- four distributed volumes on three storage nodes

#uname -a
Linux storage-node-[1-4] 5.2.5-200.fc30.x86_64 #1 SMP Wed Jul 31 14:37:17 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
#glusterfsd --version
glusterfs 6.4
--------------------------

--------------------------
client side details:

- two Ubuntu 18.04 clients are using these volumes
#uname -a
Linux 4.15.0-54-generic #58-Ubuntu SMP Mon Jun 24 10:55:24 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

#glusterfs --version
glusterfs 3.13.2
--------------------------