powering up/down the new cluster

[Francisco Chinchilla <fchinchi@richmond.edu>]

If the power went off and came back on then turn everything off first (turn 
the power switch off by hand, except on pscm1, that one you can issue a halt 
or shutdown command).  The instructions to turn the new cluster on assume 
everything is turned off and I think this is the only 'sure' way top bring it 
back up safely.

TURNING THE CLUSTER ON
1) Turn the 3 RAID machines on.  I recommend you wait 10 seconds or so before 
you turn the next one on.  You are turning on at least 10 hard drives per 
machine, and that could just pull too much power at once if you do all 3 at 
the same time.

2) Turn the actual FILESERVER on.  The fileserver will talk to the 3 RAID 
machines and lights will blink.  When this is done,

3) Turn the MASTER on.  You can turn off the sec. master on too.

4) Boot all the other cluster nodes by just turningon the powerbar switch.  
Turn on only 1 switch (4 nodes) at a time.  Turning on more than a couple 
machines at a time requires too much from the tftp server that runs on the 
master, and as a consequence they will not be able to download the kernel and 
they will all be unavailable.  The safest way is to wait till the 4 nodes that 
you just turned on become 'available' before you turn on the next one, but I 
think you can wait till they are 'up', wait 2 mins, then turn the next 4 nodes 
on.

Overall this takes at least 40 mins.


TURNING THE CLUSTER OFF
1) from the master, ssh to the fileserver and turn it off
	ssh fileserver
	shutdown -hF now
	exit (from the fileserver)
2) On the master, shutdown -hF now

3) Power down the 3 RAIDS by hitting the switch.  Now hit the power switch on 
the fileserver to really turn it off (the shutdown command just turns 
everything off so that hitting the switch is safe)

4) Power down all the strips.


NOTES
1) You can reboot the NODES with no problems, not true for the master nor the 
fileserver.  You have to go through the whole procedure again.

2) If a node doesn't become available at first, press and hold the power 
switch on that node for ~4 seconds to turn it off.  Press it again to turn it 
back on (the reset button sometimes doesnt do much, but this does).