North American Network Operators Group|
Date Prev | Date Next |
Date Index |
Thread Index |
Author Index |
RE: UUNET connectivity in Minneapolis, MN
- From: Charles Cala
- Date: Fri Aug 12 10:13:03 2005
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=lxj88n38crvxcuX3ni115eApxhb4hM7Pdf7eu674bk1E9Okkt/wfxh8rcGQLkB8HnAz6QrqFrFpR1+O9OEwfMFX5fI/VDFgiAPzHbr3j+mIN85zvm02UN2WVLFcMUpuHjoG/tU+STU7bQ6t5/w5SNZJwsfkUEcz/+08B9H9+SgU= ;
On Behalf Of James D. Butt
> Unless there is some sort of crazy story related
> to why a service provider
> could not keep the lights on, this should have not
> been an issue with
> proper operations and engineering.
6 stories from the trenches
Once a back hoe decided to punch through a high
pressure natural gas main, right outside
our offices. The fire department had us
shut down ANYTHING that MIGHT make a spark.
No nothing was able to run. It did not matter
that we had uspes and such,
all went dark for hours.
During the Northridge earthquake (the one during the
world series in sf.ba.ca.us) there was a BUNCH of
disruption of the infrastructure, drives were shaken
til they crashed, power wend down all over the area,
Telco lines got knocked down, underground vaults got
flooded, and data centers went off line.
When ISDN was king(or ya get a t-1),
I worked for an ISP in the bay area that
was one of the few to have SOME
connectivity when mae-w went down. We had a t-1 that
went “north” to another exchange point, and even
that little guy had %50+ packet loss, it kept
We were one of the few isp’s that
had ANY net connection, most of the people
went in through their local MAE ,
(that was in the days before connecting
to a MAE required that you be connected to
several other MAE’s)
Once while working for a startup in SF,
I pushed for upses and backup power gen
sets for our rack of boxes, and I was told
that we were "in the middle of the finintial district
of SF, that bart/the cable cars ran near by,
and that a big huge sub station with in
rock throwing distance of our building,
not to mention a power plant a couple
miles away. There was no reason for us to
invest in backup gen sets, or hours of
ups time…. I asked what the procedure
was if we lost power for an extended
period of time, and I was told, “we go home”
wellllllll…… the power went off to the
entire SF region, and I was able to shut
down the equipment with out to
much trouble, cause my laptop was plugged into a ups
(at my desk) and the critical servers were on a ups,
well as the hub I was on. After I verified that we
stil up at our co-lo (via my CDPD modem)
I stated the facts to my boss, and told him
that I was following his established
procedure for extended power loss.
I was on my way home. (boss=not happy)
A backup generator failed at a co-lo because
of algae in the diesel fuel.
Another time a valve broke in the buildings HVAC
sending pink gooey water under the door ,
and into the machine room.
There are reasons why a bunch of 9’s piled together,
weird stuff does happen. This is nanog, each
‘old timer’ has a few dozen of these events
they can relate.
The first 2 ya realy can’t prepare for other
than for all your stuff to be mirrored
‘some place else’, the rest are preventable,
but they were still rare.
( back to an operational slant)
Get a microwave t-2 and shoot it over to some
other building, get a freaking cable modem as
a backup, or find another way to get your lines out.
If having things work is important to you,
YOU should make sure it happens!
If people are preventing you from doing your job
(having servers up and reachable) CYA, and
point it out in the post mortem.
Curse the dark, or light a match. You decide, it's your dark.
Valdis.Kletnieks in NANOG