I first reported this on the user list, this bug report consolidates the discussion.  The full thread is available at http://gluster.org/pipermail/gluster-users/2010-January/thread.html#3719 with the subject "maxing out cpu".

It seems that using a self compiled version 2.0.9 on Ubuntu 9.10 on AMD64 the gluster client gets stuck in an infinite loop in libglusterfs/src/inode.c around lines 902-908.  This seems somewhat similar to bug #761853.

Relevant Config:
$ cat /etc/glusterfs/glusterfsd.vol 
volume posix
  type storage/posix
  option directory /srv/glusterfs/export
end-volume

volume locks
  type features/locks
  subvolumes posix
end-volume

volume brick
  type performance/io-threads
  option thread-count 4
  subvolumes locks
end-volume

volume server
  type protocol/server
  option transport-type tcp
  option transport.socket.bind-address 192.168.XXX.123
  option auth.addr.brick.allow 192.168.XXX.122,192.168.XXX.123,192.168.XXX.124
  option auth.addr.locks.allow 192.168.XXX.*
  option auth.addr.posix.allow 192.168.XXX.*
  subvolumes brick
end-volume


$ cat /etc/glusterfs/glusterfs.vol 
# Note the 3 node is missing because I removed it earlier today
volume my-storage-c
  type protocol/client
  option transport-type tcp
  option remote-host 192.168.201.123
  option remote-subvolume brick
end-volume

volume my-storage-d
  type protocol/client
  option transport-type tcp
  option remote-host 192.168.201.124
  option remote-subvolume brick
end-volume

volume replication
  type cluster/replicate
  subvolumes my-storage-c my-storage-d
  option read-subvolume my-storage-c
end-volume

volume writebehind
  type performance/write-behind
  option window-size 1MB
  subvolumes replication
end-volume

volume cache
  type performance/io-cache
  option cache-size 256MB
  subvolumes writebehind
end-volume

selective content of /etc/fstab

/dev/mapper/my--ui--c-glusterfs--export /srv/glusterfs/export ext4     noatime,nodev,nosuid 0       2
/etc/glusterfs/glusterfs.vol  /path/to/mount  glusterfs  defaults  0  0

Attached is the valgrind/callgrind output. 

The client logs contained nothing of interest.  All it had was some basic environment info and the volfile.