[ale] Running a hands off remote Linux installation

[Ed Cashin]

On Tue, Jul 31, 2012 at 2:22 PM, Jim Lynch
<ale_nospam at fayettedigital.com> wrote:
> On 07/31/2012 11:50 AM, Michael H. Warfield wrote:
> <SNIP>
>> 1) You haven't defined just what constitutes "big bucks".
> Maybe a couple of hundred including the UPS max.
>>
>> 2) You say a spare system but the problem depends HEAVILY on the
>> capability of said spare system.
> It's a tower with a late model Gigabyte MB and a 4 core AMD chip.
>>
>> * Does it support power recovery in "Always Power On"?
> I think so.  It has the option to boot when power comes back if that's
> what you mean. "The system is turned on upon the return of the AC power".

There are so many great responses in this thread, but for power
failure scenarios, you're probably most interested in avoiding
filesystem corruption or the need to fsck, so I am still wondering why
you couldn't boot the spare system into a ramfs and run a script.

After boot (external power restored after UPS exhausted), the script
would wait (e.g.) 5 minutes after boot and then telnet or ssh to the
APC to turn on the main system's power outlet.  The main system would
be pre-configured in the BIOS to always boot for power on.  The 5
minute (or whatever) delay would be long enough to allow the UPS to
charge enough to power the main system during its boot and clean
shutdown.  That would prevent on-and-off power from catching the main
system with no protection.

The serial cable from the UPS would be connected to the secondary
system.  On external power failure, the UPS would signal the secondary
system through the serial cable.  The secondary system would cleanly
shut down the main system and then turn off the main system's power
outlet at the UPS.  Because it's running from a ram disk, there's no
need to shut down the secondary system.

This setup takes three paragraphs to explain but seems very simple to
me, relying on only old and easy-to-understand technologies.  The most
inconvenient feature is the length of the delay, which would need to
be determined based on observation, or, based on your described needs,
could probably be set a bit long.

That said, I have to go read about AMT and stuff now.  :)

-- 
  Ed Cashin <ecashin at noserose.net>
  http://noserose.net/e/
  http://www.coraid.com/