[Linux-cluster] re:GFS Cookbook

[Bob Peterson]

On Fri, 2007-12-14 at 03:46 +0530, Rajagopal Swaminathan wrote:
> 
> Greetings,
> 
> apologies for reposting this I had posted this question from my yahoo
> account. Now I am a subscriber of this list
> 
> This is an adjunct question to Koustubha Kale's discussion.
> 
> I am sysadmin of the cluster Koustubha Kale had built. 
> 
> My question is:
> 
> When the node gets fenced, it is rebooted using fencing device.
> essentially let us consider that as power cycling the fenced node. Now
> this configuration has local boot disks with OS and apps loaded and
> shared storage for user home directories.
> 
> What should be done to ensure that the local storage FS is not
> corrupted? should a "touch /forcefsck" be done on the fenced node's
> local storage before it is sent on a reboot?
> 
> Perhaps I have not understood the fencing enough
> 
> Regards
> 
> Rajagopal 

Hi Rajagopal,

If your local storage is ext3, it should not have a problem.
The journals should take care of everything for you.  So it is
only the shared storage you need to be concerned about.
If that shared storage for user home directories is gfs or gfs2,
after the reboot, the file system is re-mounted, which will
replay the journals.  So again, everything should be okay.

Regards,

Bob Peterson