gull at gull.us
Wed Aug 11 23:01:54 UTC 2010
On Wed, August 11, 2010 1:18 pm, Ryan Coleman wrote:
> On Aug 11, 2010, at 3:06 PM, David Brodbeck wrote:
>> On Wed, August 11, 2010 12:25 pm, Ryan Coleman wrote:
>>> He thinks that at 500W needed it would give me about 12 minutes on a
>>> 1400VA. My consideration is, then, give the server 2 minutes on
>>> If full power has not been returned, shut down the server but leave the
>>> modem (w/ wireless) and switch running with power for up to 6 hours.
>> A bit of advice: If this is an unattended system, give some thought to
>> you will boot the server back up if the outage is longer than two
>> but shorter than six hours. Most UPS installations have *some* kind of
>> race condition issue if power comes back after the servers have begun a
>> shutdown, but in your case it's an unusually long window.
> Meaning that my 2-minute window is unusually long? If the UPS can support
> the system for 12 minutes, I say give it 20% of the life of the support
> because our power outages here are usually spikes that kill my current web
> server (but amazingly *not* my file server). In fact, one of those power
> fluxes occurred last night. I love storms for the light shows, but hate
> them for the toll they take on my servers.
Nope, 2 minutes is fine, maybe even short depending on how long your
system takes to shut down. What I'm asking about is this scenario:
1. Power goes out.
2. Server shuts itself down after 2 minutes.
3. Power comes back on before the UPS batteries are exhausted.
The server never sees a power cycle, so it doesn't boot itself back up
until someone physically goes and pushes the button.
More information about the freebsd-questions