[Linux-cluster] GFS2 hangs on one node of a two node cluster

[Marko Jung]

Hello,

this is the second time this week all my gfs2 filesystems are unaccessible 
on one of my two node cluster. I am using CentOS 5.3 (the evil AS ;-) ) and 
  the cluster otherwise works well. What is used: clvmd to manage 40TB SAN 
storage in combination with GFS2. No other repositories except Fedora EPEL5 
used.  Kernel is latest (2.6.18-128.4.1.el5) and all provided software 
updates are in place.

The issue can be described as all gfs2 filesystems being not accessible 
without any error messages by user space applications, except bash 
complaining at logins: "-bash: cd: /san/home/USERNAME: Input/output error
". Trying to unmount the filesystems also did not succeed. Just a reboot 
seems to fix the issue.

The first time the problem occurred the system got a load of >800 (!!) but 
just because all processes were waiting for IO.

Find an excerpt of the messages that seem to be related to that issue 
attached. Any hints, ideas and comments highly appreciated. Thank you for 
your good work and help!

Marko
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: gfs2-hang.log
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20090826/06d190c0/attachment.log>