Re: One more thing to watch out for at data centers - fire drills

2016-09-18 Thread Stephen Satchell

On 09/17/2016 02:43 PM, Larry Sheldon wrote:

My experiences were back in the days of washing-machine class disc
drives and they were a 4-hour fire-wall away, but I don't remember them
being impacted. (I can't believe that I was allowed to conduct a test
with them running, but I don't remember shutting them down.)

I wonder if orientation mattered--mine were all platters parallel to the
floor, I wonder if the damaged ones were parallel to the wave front.


If you watched the video of the guy who screamed at his disk drives to 
cause temporary faults, the JBOD had its platters horizontal to the floor.


One of the reason the washing-machine-sized CDC Storage Module Drives 
weren't affected by high noise level is the sheer beefy mass of the head 
assembly and the voice coil.  Also, the track spacing on the platters of 
those drives was far less dense, so any noise-induced mis-tracking would 
be minuscule, and easily handled by said voice-coil's position-error 
system.  The heads were larger, as well as the head arms.  In this 
situation, mass is your friend.


Re: One more thing to watch out for at data centers - fire drills

2016-09-17 Thread Larry Sheldon



On 9/17/2016 07:39, Suresh Ramasubramanian wrote:

http://motherboard.vice.com/read/a-loud-sound-just-shut-down-a-banks-data-center-for-10-hours?utm_source=bbcfb

 Releasing inert gas from fire suppression units that were over
pressurized resulted in an extremely loud noise


My experience is only with in-specification systems (and only in tape 
libraries) but those tests were pretty loud.


– causing cabinets
> full of hard drives to vibrate – which got transmitted to the read –
> write heads of the drives.

My experiences were back in the days of washing-machine class disc 
drives and they were a 4-hour fire-wall away, but I don't remember them 
being impacted. (I can't believe that I was allowed to conduct a test 
with them running, but I don't remember shutting them down.)


I wonder if orientation mattered--mine were all platters parallel to the 
floor, I wonder if the damaged ones were parallel to the wave front.



full of hard drives to vibrate – which got transmitted to the read –
write heads of the drives.

Amazing sort of outage + data loss, and this time the physical
security plant chief gets to write up the RCA.


--
"Everybody is a genius.  But if you judge a fish by
its ability to climb a tree, it will live its whole
life believing that it is stupid."

--Albert Einstein

From Larry's Cox account.


Re: One more thing to watch out for at data centers - fire drills

2016-09-17 Thread Ken Chase
All of these discussions sounds infinitely safe for humans.

Servers and network gear is replaceable. Sounds like the failure was not one of
DC mismanagement but human safety errors.

/kc


On Sat, Sep 17, 2016 at 09:27:26AM -0400, h...@netcases.net said:
  >On 2016-09-17 08:39, Suresh Ramasubramanian wrote:
  
>>http://motherboard.vice.com/read/a-loud-sound-just-shut-down-a-banks-data-center-for-10-hours?utm_source=bbcfb
  >>
  >>Releasing inert gas from fire suppression units that were over
  >>pressurized resulted in an extremely loud noise ??? causing cabinets
  >>full of hard drives to vibrate ??? which got transmitted to the read ???
  >>write heads of the drives.
  >>
  >>Amazing sort of outage + data loss, and this time the physical
  >>security plant chief gets to write up the RCA.
  >>
  >>--srs
  >
  >Another unexpected result when we had an all-out Halon test:  thick fog,
  >apparently from cold gas and somewhat humid air. I'm glad to have been
  >watching through windows. Visibility in the room dropped to zero.

-- 
Ken Chase - m...@sizone.org Guelph Canada


Re: One more thing to watch out for at data centers - fire drills

2016-09-17 Thread hcb

On 2016-09-17 08:39, Suresh Ramasubramanian wrote:

http://motherboard.vice.com/read/a-loud-sound-just-shut-down-a-banks-data-center-for-10-hours?utm_source=bbcfb

Releasing inert gas from fire suppression units that were over
pressurized resulted in an extremely loud noise – causing cabinets
full of hard drives to vibrate – which got transmitted to the read –
write heads of the drives.

Amazing sort of outage + data loss, and this time the physical
security plant chief gets to write up the RCA.

--srs


Another unexpected result when we had an all-out Halon test:  thick fog, 
apparently from cold gas and somewhat humid air. I'm glad to have been 
watching through windows. Visibility in the room dropped to zero.


Re: One more thing to watch out for at data centers - fire drills

2016-09-17 Thread Sascha Pollok

http://motherboard.vice.com/read/a-loud-sound-just-shut-down-a-banks-data-center-for-10-hours?utm_source=bbcfb

Releasing inert gas from fire suppression units that were over pressurized 
resulted in an extremely loud noise – causing cabinets full of hard drives 
to vibrate – which got transmitted to the read – write heads of the drives.


Amazing sort of outage + data loss, and this time the physical security 
plant chief gets to write up the RCA.


A quick comment: for the N2 systems we built in our datacenters we added 
mufflers to reduce the noise which raised the price for the system by only 
5%. Totally worth it. Yet many datacenters do not add them.


-Sascha