• machine freezes or panics under light load (pine, ghostview, cvs updates...)
• AFS access being extremly slow
• error messages like "new_VCache_entry ... failed..' in /var/log/messages

This have all been traced to the following configuration entry in /etc/sysconfig/afs

CACHESIZE=AUTOMATIC

This entry tries to guess the AFS cache size based on the assumption that a separate partition is used, and will otherwise default to 8MB. Such a small cache makes AFS run unreliably.

You can verify the current AFS cachesize via

fs getcacheparms

Workaround

Edit /etc/sysconfig/afs and set

CACHESIZE=100000

(any value larger than 64000 (64MB) should be OK, but you need to have enough space on the partition that holds /usr/vice/cache)

You will need to restart AFS afterwards, or reboot the machine.

Update

As of Feb 16, 15:00, the following commands on SLC3/i386 should be sufficent:

   /usr/sbin/apt-autoupdate -y
   /usr/sbin/lcm --update

   /usr/bin/yum update
   /usr/sbin/lcm --update

This is implemented via an update of the AFS configuration script, version ncm-afsclt-1.3.1-1.noarch runs as a "triger" after updates to openafs. Use

rpm -q ncm-afsclt