[Gluster-devel] Dropping caches causes a zombie

[Gordan Bobic]

When glusterfs is being used as rootfs, doing this:

echo 3 > /proc/sys/vm/drop_caches

makes the bash that issued the command go zombie and it cannot be killed 
or stopped. This happens on all my glfs-root servers. The only cure is 
to reboot the machine. There is no actual crash per se - the machine 
continues working and responding fine. It's particularly weird because 
it is the bash process that ends up using the CPU.

Trying the same thing on a machine that is only acting as a glfs client 
(not running root on glfs) works fine.

There are no performance translators, the brick layout is client side 
AFR with single process client and server. The OS is CentOS 5.4 on all 
boxen running kernel 2.6.18-164.9.1.el5, and glfs 2.0.9.

Gordan