[Linux-cluster] Cluster Crashes

[Lon Hohberger]

On Thu, 2006-11-16 at 17:29 -0600, isplist at logicore.net wrote:
> > Log in to the switch and check the status of the port connected to the
> > node in question.  It should be down.
> 
> I'll do that next time.
>  
> >> If a drive get's moved on my FC storage, the cluster crashes.
> >> If the storage is rebooted, the cluster crashes. If I change
> 
> > You're kidding, right?  Filesystems don't like having storage yanked
> > out from underneath them, clustered or not.
> 
> I don't mean gone as in removed, I mean changed as in adding/removing a drive
> in a system that should not cause these problems. The Xyratex gear I'm using 
> should allow me to add/remove drives while the system is running, as in hot 
> swap. I'm guessing it's the Xyratex I need to work on some more. 
> 
> Still, I do want to understand how to deal with hardware changes so that I 
> don't have crashed clusters too often since we do change/move gear in and out, 
> a lot.

Swapping a drive on a properly configured hot-swappable array should not
even be noticed by the nodes.  Worst case is that performance drops
slightly while the array rebuilds the raid set.

-- Lon