Re: [osol-discuss] Opensolaris.org is back up

2007-07-27 Thread Brian Gupta
You want enough battery time to start the generators, and fix the generators
and pumps that may possibly fail. (And you definately want to make sure you
have redundant fuel pumps for each generator.)

Also quarterly testing is a must.

--brian

On 7/27/07, Bill Sommerfeld [EMAIL PROTECTED] wrote:

 On Tue, 2007-07-24 at 22:20 -0700, Hugh McIntyre wrote:

 
 http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104STORY=/www/story/07-24-2007/0004631109EDATE=
 
  In answer to the questions about backup power, obviously they have UPS
  and generators (as the press release says...).

 There is more detail available here:

 http://www.365main.com/status_update.html

 Looks like they had  8 primary and 2 secondary generators - N+2
 redundancy.

 They lost 5 -- 4 failed to start, and another was overloaded into
 failure due to the first 4 failures.

 At first, three primary and one secondary generator failed to come up.

 It appears there was sufficient capacity in the remaining secondary
 generator to handle the load of two of the primaries, but not three, so
 the running secondary generator *also* fell over.

 As a result, three of their eight rooms lost power for about 45 minutes;
 a fourth room lost power for half a second due to some sort of PDU
 glitch.

 - Bill





 ___
 opensolaris-discuss mailing list
 opensolaris-discuss@opensolaris.org

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Re: [osol-discuss] Opensolaris.org is back up

2007-07-27 Thread Brian Gupta
One more thing. Geographical redundancy is a good thing. -brian

On 7/27/07, Brian Gupta [EMAIL PROTECTED] wrote:

 You want enough battery time to start the generators, and fix the
 generators and pumps that may possibly fail. (And you definately want to
 make sure you have redundant fuel pumps for each generator.)

 Also quarterly testing is a must.

 --brian

 On 7/27/07, Bill Sommerfeld [EMAIL PROTECTED] wrote:
 
  On Tue, 2007-07-24 at 22:20 -0700, Hugh McIntyre wrote:
 
   http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104STORY=/www/story/07-24-2007/0004631109EDATE=
 
  
   In answer to the questions about backup power, obviously they have UPS
   and generators (as the press release says...).
 
  There is more detail available here:
 
  http://www.365main.com/status_update.html
 
  Looks like they had  8 primary and 2 secondary generators - N+2
  redundancy.
 
  They lost 5 -- 4 failed to start, and another was overloaded into
  failure due to the first 4 failures.
 
  At first, three primary and one secondary generator failed to come up.
 
  It appears there was sufficient capacity in the remaining secondary
  generator to handle the load of two of the primaries, but not three, so
  the running secondary generator *also* fell over.
 
  As a result, three of their eight rooms lost power for about 45 minutes;
  a fourth room lost power for half a second due to some sort of PDU
  glitch.
 
  - Bill
 
 
 
 
 
  ___
  opensolaris-discuss mailing list
  opensolaris-discuss@opensolaris.org
 


___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Re: [osol-discuss] Opensolaris.org is back up

2007-07-27 Thread Bill Sommerfeld
On Tue, 2007-07-24 at 22:20 -0700, Hugh McIntyre wrote:

 http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104STORY=/www/story/07-24-2007/0004631109EDATE=
 
 In answer to the questions about backup power, obviously they have UPS 
 and generators (as the press release says...).  

There is more detail available here:

http://www.365main.com/status_update.html

Looks like they had  8 primary and 2 secondary generators - N+2
redundancy. 

They lost 5 -- 4 failed to start, and another was overloaded into
failure due to the first 4 failures.

At first, three primary and one secondary generator failed to come up.

It appears there was sufficient capacity in the remaining secondary
generator to handle the load of two of the primaries, but not three, so
the running secondary generator *also* fell over.

As a result, three of their eight rooms lost power for about 45 minutes;
a fourth room lost power for half a second due to some sort of PDU
glitch.

- Bill





___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-25 Thread Hugh McIntyre
Dennis Clarke wrote:
 Nothing can replace batteries.  The more the better.
 Most power solutions should grant at *least* 30 minutes on battery and if
 not then I would not host my gear there.  I just finished installing a 45
 minute battery runtime solution with 16,000 VA of power.  It is not that
 costly really to stick in an APC Symmetra SY16K unit when one looks at the
 cost of downtime.

Most datacenters consume a *lot* of power.  Just how many lead-acid 
batteries can you fit in one building?  (There are probably safety and 
weight issues, plus density problems if every other rack in your 
datacenter is a UPS).

And in any case, an outage can just as easily last an hour or more as it 
can last a few minutes.  So the usual plan is that the UPS only needs to 
last long enough for the generator to kick in, e.g. 5-10 minutes.

I've also seen people use flywheels for short term power, but again 
there's only so large a flywheel the authorities will let you install.

Hugh.

PS: at an Indian office of the company I work for, we recently managed 
to have a backup generator last for 2-3 days via continuous refueling 
because of the failure and wait-for-replacement of a step down 
transformer.  I doubt you could have had enough batteries for this...

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-25 Thread Dennis Clarke

 Dennis Clarke wrote:
 Nothing can replace batteries.  The more the better.
 Most power solutions should grant at *least* 30 minutes on battery and if
 not then I would not host my gear there.  I just finished installing a 45
 minute battery runtime solution with 16,000 VA of power.  It is not that
 costly really to stick in an APC Symmetra SY16K unit when one looks at the
 cost of downtime.

 Most datacenters consume a *lot* of power.  Just how many lead-acid
 batteries can you fit in one building?  (There are probably safety and
 weight issues, plus density problems if every other rack in your
 datacenter is a UPS).

I wasn't clear.

The batteries just give you time to start the generator.  Up here in Canada
the generator will be frozen and, possibly, the diesel fuel will be also.

Provided you have a Jeep with winter dueller tires on it and you can get to
the server room in 40 minutes then all you need is another ten or 15 minutes
to drag the jerry cans to the generator from the fuel shed which ( hopefully
) is not iced over with the locks frozen.

You see .. it can be tricky.

But cooling is real cheap.

Dennis

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-25 Thread Ian Collins
Hugh McIntyre wrote:
 Dennis Clarke wrote:
   
 Nothing can replace batteries.  The more the better.
 Most power solutions should grant at *least* 30 minutes on battery and if
 not then I would not host my gear there.  I just finished installing a 45
 minute battery runtime solution with 16,000 VA of power.  It is not that
 costly really to stick in an APC Symmetra SY16K unit when one looks at the
 cost of downtime.
 

 Most datacenters consume a *lot* of power.  Just how many lead-acid 
 batteries can you fit in one building?  (There are probably safety and 
 weight issues, plus density problems if every other rack in your 
 datacenter is a UPS).

   
The best I've seen was the central telecommunications hub of a country I
shall not mention which had 48 hours of battery backup.  That was a lot
of batteries and yes, it was on the ground floor!

Ian
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-25 Thread Tirthankar


Thanks,
Tirthankar



Stephen Lau wrote:


Dennis Clarke wrote:
 


There was a major power outage in SF that affected our servers.

 


three little letters :   U  P  S

better yet :   A P C

get some :-)
   



The whole reason we host in a datacentre is that the datacentre is 
supposed to take care of things like UPSs and generators.


Clearly, something went wrong at the datacentre level and we will follow 
up with them to make sure things are rectified for future outages.
 


They should start using the Solaris Cluster geo edition
http://www.sun.com/software/solaris/cluster/index.xml


cheers,
steve

 

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Dennis Clarke

 There was a major power outage in SF that affected our servers.


three little letters :   U  P  S

better yet :   A P C

get some :-)

Dennis
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Ché Kristo
Wow, I was under the impression it only affected residential areas...
 
 
This message posted from opensolaris.org
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Artem Kachitchkine

 Wow, I was under the impression it only affected residential areas...

I'm guessing netflix, craigslist, typepad and livejournal are not among 
your favorite sites :)

-Artem

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Manoj Joseph
Derek Cicero wrote:
 There was a major power outage in SF that affected our servers.

Discussion search does not seem to be working.

For example:
http://www.opensolaris.org/jive/search.jspa?objID=f80q=ZFS

-Manoj

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Rob Giltrap
Dennis Clarke wrote:
 There was a major power outage in SF that affected our servers.
 

 three little letters :   U  P  S

 better yet :   A P C

 get some :-)

It should have been part of their B C P ;-)


___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Dennis Clarke


 Wow, I was under the impression it only affected residential areas...

 I'm guessing netflix, craigslist, typepad and livejournal are not among
 your favorite sites :)

geez .. I think I looked at one of them once .. maybe.

as for residential areas, I have 6000 VA worth of UPS here at my desk ( at
home ) and even my X-Box has a UPS.  Never lose a game to power outage!

so why do we not have 10,000 VA worth of APC gear on the OpenSolaris.org
servers?

geez .. a few batteries are no big deal.

Dennis
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Artem Kachitchkine

 geez .. I think I looked at one of them once .. maybe.
 
 as for residential areas, I have 6000 VA worth of UPS here at my desk ( at
 home ) and even my X-Box has a UPS.  Never lose a game to power outage!

The Warlock: Why did you bring a cop to my command center?
John McClane: It's a basement!
The Warlock: Who is this man?


___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Dennis Clarke


 geez .. I think I looked at one of them once .. maybe.

 as for residential areas, I have 6000 VA worth of UPS here at my desk ( at
 home ) and even my X-Box has a UPS.  Never lose a game to power outage!

 The Warlock: Why did you bring a cop to my command center?
 John McClane: It's a basement!
 The Warlock: Who is this man?

I am not *that* much of a geek .. but .. when the power goes out I still
have lights, computers, a network switch and a television and the XBox and
..

oh no ... maybe I am ... but its not a bloody command center it is just
the basement.  Really .. its not big deal.  No no .. really.

dc

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Moinak Ghosh
Dennis Clarke wrote:
 There was a major power outage in SF that affected our servers.

 

 three little letters :   U  P  S

 better yet :   A P C

 get some :-)
   

   Heh, sometimes those three letters can themselves give up on your face:
   http://blogs.sun.com/moinakg/entry/the_belenix_0_6_ride

Regards,
Moinak.

 Dennis
 ___
 opensolaris-discuss mailing list
 opensolaris-discuss@opensolaris.org
   


___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Stephen Lau
Dennis Clarke wrote:
 There was a major power outage in SF that affected our servers.

 
 three little letters :   U  P  S
 
 better yet :   A P C
 
 get some :-)

The whole reason we host in a datacentre is that the datacentre is 
supposed to take care of things like UPSs and generators.

Clearly, something went wrong at the datacentre level and we will follow 
up with them to make sure things are rectified for future outages.

cheers,
steve

-- 
stephen lau // [EMAIL PROTECTED] | 650.786.0845 | http://whacked.net
opensolaris // solaris kernel development
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Dennis Clarke

 Dennis Clarke wrote:
 There was a major power outage in SF that affected our servers.



 three little letters :   U  P  S

 better yet :   A P C

 get some :-)


Heh, sometimes those three letters can themselves give up on your face:
http://blogs.sun.com/moinakg/entry/the_belenix_0_6_ride


on point 25 there .. I agree that APC service is excellent.

I have five units here that are 1200 VA each and then a 1500VA unit and a
few small guys kicking around.  That keeps five or six computers going as
well as some other devices.  The real trick is to do an orderly shut down
because most of these newer guys want USB connections and you just can't
plug that into the back of some of my boxes here.

the ultimate is the 48V big monster batteries I have in my server room. Two
man lift only at about 90Kg each I think.  Feels like a ton to lift. But the
run time is just amazing at well over 600 minutes.

oh well ... I see you have your fair share of APC events in your life also.

dc
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Hugh McIntyre
Ché Kristo wrote:
 Wow, I was under the impression it only affected residential areas...

More details at:
http://www.theregister.co.uk/2007/07/24/web_site_outages/

Particularly sad is the comment in the article that:

 Redenvelope continued to be unavailable, the same day PR monkeys
 issued a press release celebrating the site's two years of
 100-percent uptime at 365 Main's San Francisco facility. Oops.

Oops indeed.  For some reason the press release now produces a 404 error 
;).  Although it's available elsewhere at:

http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104STORY=/www/story/07-24-2007/0004631109EDATE=

In answer to the questions about backup power, obviously they have UPS 
and generators (as the press release says...).  But such backup systems 
only get tested rarely, and have an unfortunate tendency to not always 
work just when they are needed.  No data center can run on UPS for more 
than a few minutes even with the largest UPS, so they all need to have 
generators.

Hugh.
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Dennis Clarke

 Ché Kristo wrote:
 Wow, I was under the impression it only affected residential areas...

 More details at:
 http://www.theregister.co.uk/2007/07/24/web_site_outages/

 Particularly sad is the comment in the article that:

  Redenvelope continued to be unavailable, the same day PR monkeys
  issued a press release celebrating the site's two years of
  100-percent uptime at 365 Main's San Francisco facility. Oops.

 Oops indeed.  For some reason the press release now produces a 404 error
 ;).  Although it's available elsewhere at:

 http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104STORY=/www/story/07-24-2007/0004631109EDATE=

 In answer to the questions about backup power, obviously they have UPS
 and generators (as the press release says...).  But such backup systems
 only get tested rarely, and have an unfortunate tendency to not always
 work just when they are needed.  No data center can run on UPS for more
 than a few minutes even with the largest UPS, so they all need to have
 generators.

Nothing can replace batteries.  The more the better.
Most power solutions should grant at *least* 30 minutes on battery and if
not then I would not host my gear there.  I just finished installing a 45
minute battery runtime solution with 16,000 VA of power.  It is not that
costly really to stick in an APC Symmetra SY16K unit when one looks at the
cost of downtime.

Dennis

___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org


Re: [osol-discuss] Opensolaris.org is back up

2007-07-24 Thread Ché Kristo
correct!
 
 
This message posted from opensolaris.org
___
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org