[Linux-cluster] What is the current recommendation concerning shutting down a cluster node?

[Thomas Sjolshagen]

 On Thu, 06 Jan 2011 13:51:21 -0500, Digimer <linux at alteeve.com> wrote:
> On 01/06/2011 01:00 PM, Hofmeister, James (WTEC Linux) wrote:
>> umount gfs
>> service rgmanager stop
>> service gfs stop
>> service clvmd stop
>> service cman stop
>
> This is my method. However, note that stopping GFS unmounts the 
> volumes,
> so you can skip the manual unmount.

 Being a Fedora 14 user (at this point), rebooting the cluster node(s) 
 with shutdown -r|h now has been working fine for me since transitioning 
 to F14 (should be the same cluster stack as what RHEL 6 uses now, I 
 believe)

 Migrating any rgmanager services off of the node I'm bringing down 
 first seems to be required (httpd/apache is pretty slow at shutting down 
 and thus causes my shutdown -r generated umount of the gfs|gfs2 file 
 systems to fail)

 // Thomas

 PS: is there a specific reason why the GFS/GFS2 service shutdown option 
 doesn't do something akin to "fuser -k" for the mountpoints in order to 
 kill off any processes that would cause the umount to fail? Yes it's 
 brutal, and certain databases would be a little unhappy, but I'd argue 
 any db shutdown script that doesn't stall until the DB is actually down 
 are buggy by definition.