[Linux-cluster] Filesystem Freeze

[Lon Hohberger]

On Mon, Mar 19, 2007 at 03:42:50PM +0100, Röthlein Michael (RI-Solution) wrote:
> Hello,
>  
> We use GFS for about 1 year now on a 4 node cluster connected to a SAN. In the past weeks nearly every day we had troubles desribed here: http://www.open-sharedroot.org/faq/troubleshooting-guide/system-failures/the-whole-cluster-freezes/ as "Freeze of filesystem"
> > There you will not see anything in the syslogs. The only thing you will see is rapidly increasing load on the system an 
> > rapidly increasing amounts of processes in the D State (means waiting for I/O). Here a lock is hung and any process 
> > accessing the resource (normally directory) locked by this lock will end up in the processstate D. 

Well, yes... Each process in 'D' adds exactly 1.0 to the system load
average.

> Is there any sure way to determine which node causes the lock or any fix for this issue? 
> We use this cluster as web server running apache 2.

What release of GFS are you using, on what kernel, and what Linux
distribution?

-- 
Lon Hohberger - Software Engineer - Red Hat, Inc.