[Asterisk-Users] 911 and lawsuits and redundancy

[Steven Critchfield]

On Thu, 2004-01-08 at 10:26, Andrew Kohlsmith wrote:
> > SCSI controllers shouldn't get spooked by drive failures. Just choose
> > good controllers. This can be difficult as you find out that even
> > Adaptec has been known to have controllers that don't work well under
> > some loads in linux. Dell has a mailing list that basically is devoted
> > to the fixing of the card/driver for one of their RAID controllers.
> 
> Actually I was referring to the controller itself giving up the ghost -- not 
> losing a drive and having the controller freak out.  :-)
> 
> In the end it's all just a big game of "acceptable risk" -- In my particular 
> situation I can handle being down for an hour while I hack together a 
> replacement system, so redundant power supplies aren't an issue.  Redundant 
> disks are but if the controller fries it's no biggie.

Agreed about the acceptable risk. So what is acceptable is defined a bit
by recovery time and effort. I have 2 asterisk machines that recovery
time must absolutely be minimal. Luckily one machine has no special
hardware, and the software and configs are routinely backed up. The one
with special hardware(T400P) can be replaced quickly. I have a T100P at
home that I can use and the machine in the office has a T100P in it.
This would get our minimal port density we need at the moment. 

I would say that redundant PSUs are important though. I can point out a
time when my company had to do maintenance to our rack hardware. With
all the machines in the rack enjoying dual PSUs, we where able to
carefully remove the machines from the rack and reroute power away from
the rack while all machines stayed up and running. We then where able to
reinstall the machines in the rack after maintenance again without ever
powering down the machines.

While I wouldn't suggest anyone do what we did, nor suggest you buy
redundant PSUs just so it is possible to do what we did, but that was a
side benefit. The whole Idea of putting your switch on redundant UPSs
also is nice as you don't always know your UPS will behave until it is
tested. 

Another similar note from experience. We have a large UPS with 5
external sealed batteries almost car sized. While moving some stuff
around the area where that UPS was located, the connector for the
external batteries became disconnected. It wasn't till we had a power
failure did we notice the UPS couldn't handle the load we had placed on
it. Had we used the dual PSUs on our office phone switch on different
circuits then we wouldn't have lost the phones. But in this case, we did
have a failure. One failure showed 2 problems. In our case we had just
popped a breaker and lost power to a wall of our office. That in itself
is poor electrical planning. This failure pointed out our UPS wasn't
properly set up to handle the load for more than a few minutes without
the external batteries.

Anyways, while hotswap PSUs may not be important, any machine that is
important is important enough for redundant supplies. It doesn't add
much to the cost of the case and is a good insurance policy. 
-- 
Steven Critchfield  <critch at basesys.com>