UPS suggestions
-bill-
It seems more expensive to replace than to buy a new one with shutdown
software.
Anyone have any suggestions of brands to look at and/or avoid.
This is for a 5.04 system with 300 watt power supply, and 17 inch color
monitor.
--
-bill-
Technical Service Systems - bi...@TechServSys.com
KG
cheaper to just buy a new one. I had a Best Power Products UPS that I
bought in 1991 for over $7000. It was hard-wired into the computer
room, and every outlet in there was on battery backup. It is still
running though, even though the battery life was supposed to be only 3
years. However, today, even though I can recommend Best highly, I would
probably go with a regular APC [American Power Conversion] UPS that are
available from many mail order houses, and also from Office Depot.
--Keith
Ken Wolff
------------------------------------------------------
Ken Wolff
MAXIMUS
Child Support Collection Center
http://www.maxinc.com
Phone: 616-957-4949 Ext: 111
Toll Free: 800-722-2338
FAX: 616-957-1614
------------------------------------------------------
Jeff Liebermann
>it is time to replace our server's ups batteries.
>It seems more expensive to replace than to buy a new one with shutdown
>software.
Huh? Can I have your old UPS with the dead batteries?
See:
http://www.apcc.com/support/battery.cfm
for replacement APC batteries. Most of the parts distributors carry
lead-acid battries. See:
http://info.digikey.com/EC/V3/493.pdf (near the bottom)
I get mine from various "surplus" dealers at somewhat lower prices.
>Anyone have any suggestions of brands to look at and/or avoid.
>This is for a 5.04 system with 300 watt power supply, and 17 inch color
>monitor.
In the size you're look for, methinks APC SmartUPS series would be the
best choice. The SmartUPS series puts out a good approximate sine wave,
while the BackUPS puts out a stepped mess. I played with Tripp-Lite and
think they're junk. If you're serious about running on backup, then be
sure to select a UPS that will function when fed with a gasoline engine
generator.
I'll spare you my lecture on why 117vac SPS's (Stand-by Power Systems)
are a poorly conceived band-aid. The server power supply should run
directly off a 24VDC or 48VDC battery pile and not require a switch from
commerical to backup power.
--
Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
(831)421-6491 pgr (831)426-1240 fax (831)336-2558 home
http://www.cruzio.com/~jeffl WB6SSY
je...@comix.santa-cruz.ca.us je...@cruzio.com
Bill Vermillion
wrote:
Just be aware that there are different models in all the lines and
the last time I looked there were two 'classes' of UPS from APC.
One was a true UPS - on line all the time, the other was more like
the typical PC standby UPS. The former is the best way to go -
but it is the most expensive.
The BEST line originally had only true UPS devices, but then
developed a lower priced line to compete with the other entries.
--
Bill Vermillion bv @ wjv.com
Scott Roberts
Jeff Liebermann wrote in message
<377dfc2b...@news.ricochet.net>...
>I'll spare you my lecture on why 117vac SPS's (Stand-by Power
Systems)
>are a poorly conceived band-aid. The server power supply should run
>directly off a 24VDC or 48VDC battery pile and not require a switch
from
>commerical to backup power.
I'm sure this isn't the place, but is there a power backup newsgroup?
We use APC BackUPS 400s in our stores which serve us fine for day to
day use. However, when the tornado took out an entire store, the
entire area was without power (we're in OKC). We tried to bring
another BackUPS from the office (the one in the affected store was
already run down) to open our electronic safe, but it doesn't supply
power unless it is already plugged in to a live outlet when you turn
it on. If it's not plugged in when you turn it on, you get no power
even though the battery is charged. Is that what you mean by a SPS?
What are the alternatives and is there a significant price difference?
And, what about running off a generator (we had several stores doing
this for a few days, but the BackUPS didn't like it)?
Scott
BTW, the computer (an Acer running DOS) survived, data in tact. Pretty
amazing if you saw the store (or the remains of it).
Bill Vermillion
<je...@comix.santa-cruz.ca.us> wrote:
>>it is time to replace our server's ups batteries. It seems more
>>expensive to replace than to buy a new one with shutdown software.
>Huh? Can I have your old UPS with the dead batteries? ...
>I'll spare you my lecture on why 117vac SPS's (Stand-by Power
>Systems) are a poorly conceived band-aid.
Maybe you need to retreive that from some old post and put it up on
your site. This is an area where many people think they are
protected, and often are deluding themselves.
>The server power supply
>should run directly off a 24VDC or 48VDC battery pile and not
>require a switch from commerical to backup power.
I remember seeing an AT&T 3B20D - with several HUGE truck
batteries. Fully redudant - everything including CPUs could be
swapped with power on. Design target was 1 hour down-time in a 20
year life-span. Almost all reach that goal - under 5 min year.
They are/were integral parts of the ESS5 switches that keep the
phones running.
Jeff Liebermann
<sgro...@nospam-ionet.net> wrote:
>I'm sure this isn't the place, but is there a power backup newsgroup?
I couldn't find one but that doesn't mean that it doesn't exist. The
closest approximation I could find is:
alt.energy.homepower
comp.sys.ibm.pc.hardware.misc
>If it's not plugged in when you turn it on, you get no power
>even though the battery is charged. Is that what you mean by a SPS?
It's a feature, not a bug. SPS's have GFI (ground fault interrupter)
circuits that prevent the beast from killing you when you lift the safety
ground or where there is any indication of leakage currents. Having been
launched by a miswired power strip, I can personally appreciate this
feature.
SPS means "Standby Power Supply" which means it's not generating power
all the time. When the AC power fails, it turns on. Otherwise, it's in
standby mode.
UPS means "Uninterruptable Power Supply" which really means that it's
generating power all the time. In a real UPS, your computer would be
running off the power generated by the UPS. The commercial AC power
would just act as a battery charger. The big advantages are zero
switching time and fantastic isolation from garbage on the power lines.
The disadvantages are cost, noise, heat, RFI, etc.
>What are the alternatives and is there a significant price difference?
>And, what about running off a generator (we had several stores doing
>this for a few days, but the BackUPS didn't like it)?
The BackUPS series are the APC cheapo series. Most will not run from a
dirty waveform or generator. The SmartUPS series are more expensive, put
out a sine wave, and most important, will run from a generator. For long
term operation, there's nothing like a generator for low cost and run
time.
There is one other alternative. You can build your own fulltime power
system that only requires a UPS for the monitors. See:
http://www.icpacquire.com.tw/ace925c.htm
http://www.icpacquire.com.tw/acer20a.htm
for 12VDC, 24VDC and 48VDC mini-tower power supplies in various
configurations. Your computers run off the batteries directly.
Commercial power just keeps the batteries charged. I have a few
installations that I've retrofited with this arrangement. Some of my
industrial control installations use DC power. It's also common in the
telco industry which runs entirely on -48VDC.
>BTW, the computer (an Acer running DOS) survived, data in tact. Pretty
>amazing if you saw the store (or the remains of it).
Ouch. Disasters are no fun. In the earthquake of 1989, I lost a few
hard disks because the servers were on a UPS. The UPS kept the drives
spinning while the ground kept shaking. Eventually, the heads bounced
off the platter killing the drives. If the UPS had shut off at the first
sign of a quake, the drives would have survived. There's a moral here,
but don't ask me what it should be.
MrSmiley
|it is time to replace our server's ups batteries.
|It seems more expensive to replace than to buy a new one with shutdown
|software.
|
|This is for a 5.04 system with 300 watt power supply, and 17 inch color
|monitor.
APC Smart UPS 1000 Net was my choice after Tripp Lite proved to be
very poor in supporting OpenServer 5. I highly recommend APC,
and that ups has enough juice for 2 pc's, and is very flexible.
Very nice shutdown software, though it must be attached
using their serial cable to tty1 or tty2, not to a serial expansion
bus card setup or it won't go all the way down.
mr:)
Scott Roberts
Jeff Liebermann wrote in message
>Ouch. Disasters are no fun. In the earthquake of 1989, I lost a few
>hard disks because the servers were on a UPS. The UPS kept the
drives
>spinning while the ground kept shaking. Eventually, the heads
bounced
>off the platter killing the drives. If the UPS had shut off at the
first
>sign of a quake, the drives would have survived. There's a moral
here,
>but don't ask me what it should be.
It's this: no matter what you do, it is wrong. :-)
Thanks for the info.
Scott
Bill Vermillion
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>Ouch. Disasters are no fun. In the earthquake of 1989, I lost a
>few hard disks because the servers were on a UPS. The UPS kept
>the drives spinning while the ground kept shaking. Eventually,
>the heads bounced off the platter killing the drives. If the UPS
>had shut off at the first sign of a quake, the drives would have
>survived. There's a moral here, but don't ask me what it should be.
The moral is more sensors. Mount something like a mercury switch
to a secure part of the wall/building and carefully set it.
In addition to the UPS monitoring the power, you could also
monitor temperature and invoke shutdowns on events there too.
The big problem is that people make assumption.
The assume that if they have a UPS their entire system is
protected.
The also often assume that locking backups in a fire-proof safe
is okay so they don't have off-site backups.
The first breaks when anything connected to the computer is plugged
into something other than the UPS. For real fun plug them into
separate legs of the power supply with marginal grounds.
The second assumption is that the disaster will be cause by
fire. If there is a fire and the safe is fire-proof - is the safe
also water-proof?
Most people seem to automaticaly think of fire. They really need
to think along the lines of "what if I can't get anything from the
building/site" - and plan for that.
Not data related - but a great many remember from about 8-10 years
ago something that happened here and made news around the world as
it was so different everyone had to see it. It was about 2 miles
from my house.
A woman noticed that here orange tree was sinking. Then it went
down, her house followed, 1/2 of a city municipal swimming pool,
and three brand new Prosches followd the tree into the sinkhole.
Data in a fireproof safe - as heavy as safes are - would probably
be at the bottom of the heap and never seen again.
Bill
JeffG
hurricanes. I removed the internal battery from my APC 600 and
connected the UPS battery wires to an 8D, 225 amp/hr Gell Cell Marine
Deep Cycle battery. I have run my system for many hours on this
battery. I use a 20 amp Stat Power smart/digital battery charger to
charge the Gell Cell. When the power is off for extended periods I use
a Honda generator to power the battery charger to keep the battery
charged. This way computer never sees current from the generator.
Jean-Pierre Radley
| A woman noticed that here orange tree was sinking. Then it went
| down, her house followed, 1/2 of a city municipal swimming pool,
| and three brand new Prosches followd the tree into the sinkhole.
|
| Data in a fireproof safe - as heavy as safes are - would probably
| be at the bottom of the heap and never seen again.
At my age, it's getting harder and harder to pump fresh helium into the
balloons tethered to the railing on my terrace which float my backup
tapes 600 feet above my apartment, which is already on the fifteenth
floor. But the fluttering tapes provide me with so much delight,
buffeted about by aeolian agents.
Do you think my tapes are getting so much X-ray or cosmic-ray or
ultra-violet exposure as to render them questionable?
--
Jean-Pierre Radley <j...@jpr.com> XC/XT Custodian Sysop, CompuServe SCOForum
Bill Vermillion
Turning serious for a moment - the two main things that damage tapes
are high-humidity and heat. As far as accidentally erasing them
goes they need to be near a very high field - eg setting on top of
a 20HP or more electric motor for example.
From a study reported in a tape care booklet published by 3M a long
time ago - there was an interesting statistic on what damages tape.
They used a unit of measure, whose name I have forgotten (was it
RAD?) that will kill 1/2 of the population exposed to it.
It took a level 500 times that amount to do the same amount of
damage to tape as 1 hour at 150F.
Cool and dry is the answer. So just as long as you have little
sunshield over the tape you should be fine. Bring them in when it
rains.
Of course if you have noted all the 'action' films on television,
you'd apt to worry about low-flying helicopters.
-bill-
Thanks
-bill-
The big problems with 48v power supplies as I see it are:
cost
needing special ventilation for the batteries/special battery room
a 48 v battery pile represents a large danger.
-bill-
> sign of a quake, the drives would have survived. There's a moral here,
> but don't ask me what it should be.
The moral is: when the earth shakes, immediately drive to all you
clients' offices and pull the plugs.
Tom Ayers
to properly shut down the system. If would need to cut power instantly,
however the platters would continue to spin down for a few seconds, and
could be damaged then. As for the sudden loss of power, what about the
smaller quakes (and/or aftershocks). What if someone bumped the
sensor? Sounds like more risk than benefit.
Anyone who is in a seismic risk area needs to think about their off-site
backup plan. I lost my home in the '94 Northridge quake. Along with it
went the off-site backup tapes. As my job was closer to the epicenter,
I was a bit concerned, fortunately, the server survived the quake, along
with the backup tapes in the fire-proof safe.
We now send a second copy of the backup to one of our sites, over 1000
miles away.
Tom Ayers
Bill Vermillion wrote:
>
> In article <377e46b7...@news.ricochet.net>,
> Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>
> >Ouch. Disasters are no fun. In the earthquake of 1989, I lost a
> >few hard disks because the servers were on a UPS. The UPS kept
> >the drives spinning while the ground kept shaking. Eventually,
> >the heads bounced off the platter killing the drives. If the UPS
> >had shut off at the first sign of a quake, the drives would have
> >survived. There's a moral here, but don't ask me what it should be.
>
> The moral is more sensors. Mount something like a mercury switch
> to a secure part of the wall/building and carefully set it.
> In addition to the UPS monitoring the power, you could also
> monitor temperature and invoke shutdowns on events there too.
>
> The big problem is that people make assumption.
>
> The assume that if they have a UPS their entire system is
> protected.
>
> The also often assume that locking backups in a fire-proof safe
> is okay so they don't have off-site backups.
>
> The first breaks when anything connected to the computer is plugged
> into something other than the UPS. For real fun plug them into
> separate legs of the power supply with marginal grounds.
>
> The second assumption is that the disaster will be cause by
> fire. If there is a fire and the safe is fire-proof - is the safe
> also water-proof?
>
> Most people seem to automaticaly think of fire. They really need
> to think along the lines of "what if I can't get anything from the
> building/site" - and plan for that.
>
> Not data related - but a great many remember from about 8-10 years
> ago something that happened here and made news around the world as
> it was so different everyone had to see it. It was about 2 miles
> from my house.
>
> A woman noticed that here orange tree was sinking. Then it went
> down, her house followed, 1/2 of a city municipal swimming pool,
> and three brand new Prosches followd the tree into the sinkhole.
>
> Data in a fireproof safe - as heavy as safes are - would probably
> be at the bottom of the heap and never seen again.
>
> Bill
Jeff Liebermann
>I'll take the lecture.
OK. However, you get an abbreviated version or I'll be late for what
will hopefully be a free lunch. Unfortunately, the door price is a box
of RTL8029 based ethernet cards. Yech.
>The big problems with 48v power supplies as I see it are:
> cost
> needing special ventilation for the batteries/special battery room
> a 48 v battery pile represents a large danger.
Yep. However, have you ever seen how really big UPS's are setup? They
don't just drop some car batteries on the floor and run clip leads to the
equipment. The battery vents are on a manifold, the equipment cabinet is
desiccated and explosion proof, and the whole mess complies with
applicable NEC codes. There are fuses, cut-outs, fire detectors, and
instrumentation that can be added. It's not that hard but you can't do
it in the middle of an office or in most server rooms. I also would not
attempt to build a large DC operated system without professional and
experienced help. As a general rule, you can build a box around gel
cells, but wet cells should be desiccated and ventilated.
For small installations (<3 machines), the batteries can be sealed gel
cells on a palette. Basically, it's similar to solar alternative energy
installation. A good fast fuse will protect the battery pile from the
inevitable screwdriver across the terminals. Out gassing from dessicate
gel cells is minimal and is not a problem in a large enough room. I
would not do it in an unventilated closet mostly because of the heat
build up, and not the hydrogen accumulation. Almost any kind of
protective enclosure will suffice. Look at the way the larger Best and
APC SPS's are packaged for a clue.
Where I get into trouble is when I have a growing installation that
starts out small and adds a new server every few months. I had one
customer with 6ea DC operated servers running on a bank of 24ea 2volt
Trojan batteries charged by an enormous fork lift battery charger. I did
not design this mess but will confess to being involved in making it
work. Were it not in a well ventilated warehouse, I would be worried. I
successfully lobbied against the aesthetics committee's desire to enclose
it with a wall for safety reasons.
I also use DC power on mountain top radio sites. Besides the completely
unreliable power, there are generator limitations at some sites. Even
with a generator, the computers tend to be more sensitive to power
glitches than the radios. Since many sites run on solar power anyway, DC
powered computers are an obvious expedient. However, that's not what
inspired me to use DC. It was because the common SPS product of several
years ago was not terribly RF proof. My logs showed false tripping,
erratic operation, and uncontrolled output that I suspected was caused by
the high RF fields. The solution was to simply remove the SPS from the
picture.
Another wonderful surprise is that according to the EPA, anyone storing
more than about 500 lbs of batteries is now the proud owner of a
hazardous material and toxic waste dump and therefore must comply with
environmental regulations. This is largely being ignored with clean
installations, but I slammed into the EPA when one customer made a huge
messy pile of dead batteries, in plain sight, and let them sit for 6
months. Eventually someone complained and the EPA arrived in force. My
next car will NOT be another pickup truck.
Let's do some math and guesswork. The typical server uses about 170
watts (1.5amps) at 117VAC. This does NOT include the monitor which will
run off the usual APC UPS. 170 watts at 24VDC is 7 amps. I use 2ea deep
cycle type 27 marine gel type batteries rated at about 100 Amp-hours. I
can usually run these down to 75% of capacity without damage and 60% if I
don't care about battery lifetime. 75% of 200 A-Hr is 150 A-Hr or 21
hours run time for a single server that uses 7 amps. I've run tests and
this is roughly what I get. Now, try to buy a 117VAC SPS that will run a
server for 20 hours for under $2000. Also, with such a long runtime, I
don't bother with automatic shutdown scripts.
I'll spare you my list of home power related URL's. You should be able
to find lots on alternative power generation. The safety issues are
often discussed.
The biggest headache is finding a suitable battery charger. The
Statpower 12volt incantations in the marine supply and altenative energy
catalogs are fine. However, I want 24V and 48VDC and these are difficult
to find in low power versions.
I'm late, more later if anyone is interested...
Robert Steinmetz
> I'll take the lecture.
>
> The big problems with 48v power supplies as I see it are:
> cost
> needing special ventilation for the batteries/special battery room
> a 48 v battery pile represents a large danger.
>
> --
>
> -bill-
>
> Technical Service Systems - bi...@TechServSys.com
There is one other problem many people who have not worked with these
usually forget ;
THEY ARE VERY HEAVY
Jean-Pierre Radley would need those balloons on the fifteenth floor, or
he would be in the basement.
--
Robert Steinmetz, AIA email: r...@steinmetznet.com
Principal
Steinmetz & Associates
Bill Vermillion
>> I'll take the lecture.
>> The big problems with 48v power supplies as I see it are: cost
>> needing special ventilation for the batteries/special battery
>> room a 48 v battery pile represents a large danger.
>There is one other problem many people who have not worked with these
>usually forget ;
>THEY ARE VERY HEAVY
The battery cabinet at one site - where I helped them move it in -
weighed about 1800 pounds.
Jeff Liebermann
<t...@SPAMBEGONE.polyflow.com> wrote:
>Earthquakes happen too fast. A shaking sensor wouldn't have enough time
>to properly shut down the system. If would need to cut power instantly,
>however the platters would continue to spin down for a few seconds, and
>could be damaged then.
The 1989 earthquake kept the ground shaking for months. It wasn't just
one big bang or collection of big bangs. It was 20 seconds of horizontal
thrashing, following by months of almost continuous minor shaking.
Long ago, the drive manufacturers discovered that the cause of stiction
and head clogging was landing the heads on the platter. The obvious
solution was to implement a head lifter. I've performed autopsies on
dead drives and found all the later incantations to contain some sort of
head lifting or land mechanism. Kill the power and the head is safe.
IMHO, what killed the drive was not the initial 20 seconds of 6.9 richter
rock and roll, but 15 minutes of continuous pounding until the UPS quit.
It was 15 minutes because brain dead (me) decided that if the UPS could
run for 30 minutes, a 15 minute delay in doing a graceful shutdown was
about right. However, even that may not have been the killer as the
drive continued to spin merrily for several hours until the UPS finally
gave up. The system was sitting there doing the "hit any key to reboot",
the drive was still spinning, and the head was still pounding on the
platter for hours.
Given the choice between attempting a graceful 5 minute "please logout"
ceremony versus a kill the power immediately crash, I'll take the crash.
Sure, some data will probably be lunched and some writes may be
incomplete, but OSR5 has a journaling filesystem which does a truly
impressive job of keeping the filesystem intact. I've killed the power
on my OSR5 box more than once and never lost a thing. (Of course, my
OSR5 box does very little so there's no risk).
I was seriously considering some kind of quake detector shutdown for my
UPS/SPS systems. However, I never bothered. Initially, I thought a tilt
detector was the right way. However, most of my servers are currently
strapped to bakers racks and will not fall over.
>As for the sudden loss of power, what about the
>smaller quakes (and/or aftershocks). What if someone bumped the
>sensor? Sounds like more risk than benefit.
Right. Someone is going to pick up a 200lb SPS and drop it while it's
running a server. However, earthquakes have clearly defined frequency
components which can be detected by a suitable low frequency
accelerometer. It would be fairly easy to distinguish between a quake
and an accidental bumping. More serious is how to test if it works.
"Kick here to test" or something equally disgusting?
>Anyone who is in a seismic risk area needs to think about their off-site
>backup plan.
Yep. My office computer, backup tapes and cash box were in an "unsafe"
building. To insure that I would not sue the People's Republic of Santa
Cruz, everyone was denied access to the downtown area, while newspaper
reporters, photographers, celebrities, street people, looters, CYA
volunteers, police, fire, inspectors, and officials wandered around
without much concern. I eventually had to burglarize my own office to
grab my backup tapes and cash box. After one court order, a threat of
organized protest, and a few weeks of begging, the bureaucracy decided to
allow 3 people per office, to empty their offices for exactly 3 hours, in
shifts of about 450 people. The traffic from 150 trucks arriving and
leaving simultaneously was amazingly efficient. Needless to say, it was
not enough time for everyone. I watched in horror as one group threw
computers and equipment out a 2nd story window because it was obvious
that they were not going to meet the 3 hour deadline. The amazing part
of all the confusion was that nobody bothered to clean up the debris from
the streets or sidewalks and nobody remembered to bring a broom. I guess
lots of debris looks good for the photographers that were all over the
place.
Try this mental exercise. An earthquake just trashed your office. You
have exactly 3 people and 3 hours to rescue what you can. Good luck. Oh
yeah, no heavy equipment, no electricity, no elevators, no lights, debris
all over the sidewalks, and two people on the stairs at a time due to
fear of collapse.
Meanwhile, some of my more careful customers had a different problem.
They had placed their backup tapes in their safe deposit boxes. It was a
week before the banks could open again and a few more days before anyone
could get to their safe deposit boxes. The moral here is to not put your
backup tapes anywhere within the authority of officials, bureaucrats, and
paranoids.
>I lost my home in the '94 Northridge quake.
Ouch. I saw that mess and it looked bad. I have my tapes in a safe at
the house. Methinks it could withstand most disasters. We don't have
sink holes in California.
-bill-
> The battery cabinet at one site - where I helped them move it in -
> weighed about 1800 pounds.
>
> My next car won't be a pickup!
You guys are full service consultants !
Tom Ayers
>
> On Thu, 01 Jul 1999 09:28:06 -0700, Tom Ayers
> <t...@SPAMBEGONE.polyflow.com> wrote:
>
> >Earthquakes happen too fast. A shaking sensor wouldn't have enough time
> >to properly shut down the system. If would need to cut power instantly,
> >however the platters would continue to spin down for a few seconds, and
> >could be damaged then.
>
> The 1989 earthquake kept the ground shaking for months. It wasn't just
> one big bang or collection of big bangs. It was 20 seconds of horizontal
> thrashing, following by months of almost continuous minor shaking.
>
> Long ago, the drive manufacturers discovered that the cause of stiction
> and head clogging was landing the heads on the platter. The obvious
> solution was to implement a head lifter. I've performed autopsies on
> dead drives and found all the later incantations to contain some sort of
> head lifting or land mechanism. Kill the power and the head is safe.
>
> IMHO, what killed the drive was not the initial 20 seconds of 6.9 richter
> rock and roll, but 15 minutes of continuous pounding until the UPS quit.
> It was 15 minutes because brain dead (me) decided that if the UPS could
> run for 30 minutes, a 15 minute delay in doing a graceful shutdown was
> about right. However, even that may not have been the killer as the
> drive continued to spin merrily for several hours until the UPS finally
> gave up. The system was sitting there doing the "hit any key to reboot",
> the drive was still spinning, and the head was still pounding on the
> platter for hours.
>
Our quake was similar. The day of the quake, I don't think the shaking
ever stopped. As the days went by, it calmed down quite a bit. I tend
to think in terms of structural damage, which was done in about the
first 5 seconds (since that is what I experienced), but it makes sense
that constant battering would do more damage to a drive.
I had my UPS set to start shut down at 5 min, and cut the AC power at 10
min. We didn't have commercial power back to the office for 7 days, so
my drives were off (and NOT spinning) during the first week of shaking.
Perhaps that is why they survived.
> Given the choice between attempting a graceful 5 minute "please logout"
> ceremony versus a kill the power immediately crash, I'll take the crash.
> Sure, some data will probably be lunched and some writes may be
> incomplete, but OSR5 has a journaling filesystem which does a truly
> impressive job of keeping the filesystem intact. I've killed the power
> on my OSR5 box more than once and never lost a thing. (Of course, my
> OSR5 box does very little so there's no risk).
>
> I was seriously considering some kind of quake detector shutdown for my
> UPS/SPS systems. However, I never bothered. Initially, I thought a tilt
> detector was the right way. However, most of my servers are currently
> strapped to bakers racks and will not fall over.
>
> >As for the sudden loss of power, what about the
> >smaller quakes (and/or aftershocks). What if someone bumped the
> >sensor? Sounds like more risk than benefit.
>
> Right. Someone is going to pick up a 200lb SPS and drop it while it's
> running a server. However, earthquakes have clearly defined frequency
> components which can be detected by a suitable low frequency
> accelerometer. It would be fairly easy to distinguish between a quake
> and an accidental bumping. More serious is how to test if it works.
> "Kick here to test" or something equally disgusting?
I was thinking in terms of the constant aftershocks. At what point do
you cut the power? I would have a lot of upset users if I cut the power
every time we had an aftershock.
After three days, I decided to "loot" my own house, which I was not
allowed to go in (red tagged). If I had waited, someone else might have
done it for me. Your city was probably also worried about looters, so
everyone had to stay away.
> >I lost my home in the '94 Northridge quake.
>
> Ouch. I saw that mess and it looked bad. I have my tapes in a safe at
> the house. Methinks it could withstand most disasters. We don't have
> sink holes in California.
Is your safe waterproof (without a fire)? A lot of fire safes become
waterproof in a fire, because the plastic casing melts together and
seals the safe, but if you don't burn, they leak at the lid. Along with
the structural damage, of course I broke a few pipes. The safe
survived, but the water is what did my tapes in. Not only did my safe
leak (at the lid), but it didn't drain out, so everything in the safe
was sitting in water for three days.
The moral is that no matter how perpared you think you are, Murphy will
find a way to prove to you that you arent.
Tom Ayers
Jeff Liebermann
<t...@SPAMBEGONE.polyflow.com> wrote:
>I had my UPS set to start shut down at 5 min, and cut the AC power at 10
>min. We didn't have commercial power back to the office for 7 days, so
>my drives were off (and NOT spinning) during the first week of shaking.
>Perhaps that is why they survived.
A fast reply due to lack of time.
*ALL* of the laptop 2.5" drives have a G force sensor that detects
exessive accelleration and parks the heads. If they didn't, the drives
would never survive. I have no idea if server class (10000 RPM) drives
have the same feature. Probably not. However, if this were available,
the need to kill the power during a quake would be eliminated.
>I was thinking in terms of the constant aftershocks. At what point do
>you cut the power? I would have a lot of upset users if I cut the power
>every time we had an aftershock.
I don't know. You're correct that there are some decisions to be made
here. Methinks a time delay would be best as the aftershocks tended to
be very short lived (<2 seconds) while the initial quake lasted 20
seconds. A 5 second delay should be adequate to prevent a false
shutdown.
Most of the real damage in my customers server caves was caused by the
sprinkler system belching rusty water all over the place. Nothing like
walking into the closet to find floppies floating in a puddle. Some of
the servers shut down because the rusty water shorted the power supplies
and not because of any fail safe graceful shutdown. At the time (1989)
few UPS's had a power off mode. Alarm outputs consisted of "I'm running"
and "I'm not".
>After three days, I decided to "loot" my own house, which I was not
>allowed to go in (red tagged). If I had waited, someone else might have
>done it for me. Your city was probably also worried about looters, so
>everyone had to stay away.
I'll spare you the looter horror stories. Suffice that draconian
measures were initially applied to insure security and then forgotten a
week later. Most of the looting was by the building occupants and
owners. The police generally ignored them until one local bozo decided
to flaunt the "legal looting" in their face by plastering the story all
over the newspapers. He got his office cleaned out but resulted in
making it more difficult for anyone else to do the same.
>Is your safe waterproof (without a fire)?
Ummmm, nope. It's welded steel with a 2" fiberglass wool insulation.
>A lot of fire safes become
>waterproof in a fire, because the plastic casing melts together and
>seals the safe, but if you don't burn, they leak at the lid.
I keep my working backup tapes in those at all my customer sites. It's
everything I can do to get them to keep the lid locked. If it's not
locked, it will fly open during a fire or quake.
>the structural damage, of course I broke a few pipes. The safe
>survived, but the water is what did my tapes in. Not only did my safe
>leak (at the lid), but it didn't drain out, so everything in the safe
>was sitting in water for three days.
Yep. I've seen that after a fire. In 1989, Microsloth replaced every
single orginal label floppy disk that was ruined. This was a big company
and they had LOTS of disks. My guess is about 500 floppies were replace
at no charge. One point for Microsloth. For the others, I baptized the
floppies first in clean water followed by a 50% alcohol-water mix. They
dried in about 2 days. The 3.5" floppies did quite well with this
treatment, but the toilet paper friction pad in the 5.25" floppies tended
to wrinckle or disolve. Only the ones that we tried to use that were
still full of dirt or rust were ruined.
>The moral is that no matter how perpared you think you are, Murphy will
>find a way to prove to you that you arent.
Yeah, I know. However, I keep trying. DC UPS's are one small step in
the right direction. I don't have time now, but if anyone is interested,
the cable tv companies are looking into flywheel operated UPS's for their
backup power. I've had some experience with these and they're really
neat. If you're worried about the damage an outgassing battery can do,
wait until you see the havoc that a runaway flywheel can cause. That's
why they bury flywheel UPS's.
More power to you...
Bill Vermillion
Tom Ayers <t...@SPAMBEGONE.polyflow.com> wrote:
>Is your safe waterproof (without a fire)? A lot of fire safes
>become waterproof in a fire, because the plastic casing melts
>together and seals the safe, but if you don't burn, they leak at
>the lid. Along with the structural damage, of course I broke a few
>pipes. The safe survived, but the water is what did my tapes in.
>Not only did my safe leak (at the lid), but it didn't drain out, so
>everything in the safe was sitting in water for three days.
A pain but not a total disaster. Wet tape can be recoverd if
you take enough time and don't try to force things.
In the audio/video field there were problems with the polyurethane
binders needed for polyvinulchoride tapes. It was hygroscopic -
that means it sucks the water out of the air.
On slow speed device - 15 and 30 ips audio tape - the tape would
stick to the head - and sometimes stretch and break if you didn't
catch it in time. In video - where depending on the format the
effective linear tape speed is between 200 and 1000 ips, it can be
instant disaster.
For audio you can bring the tape up to about 120 degress farenheit
- NO MORE THAN THAT - and slowly bake. The molecule changes formed
by the moisture in the binder is reversible.
Storing for weeks/month in a .5% to 1.0% RH (relative humidity)
environment will also take care of it.
In data tape you typically won't have binder problems and if the
tape is left wound the interior of the tape will only become wet
via capillary action.
On videotape - particular those in cartridges such as beta, vhs,
and u-matic, restoration houses typically freeze the tapes in
commercial freezers - it needs to be in the -40F range.
This effectively freeze-drys the tape in the way that is quite
similar to freeze-dried food in a grocery store.
It boils down to the choice of is it cost effective for a data
recovery service to retreive the data, or is it cheaper just
to re-enter, or update from older backups.
A great many disasters aren't non-recoverable, just difficult.
Pounding heads on an HD is quite another story.
Bill Vermillion
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>*ALL* of the laptop 2.5" drives have a G force sensor that detects
>exessive accelleration and parks the heads. If they didn't, the drives
>would never survive.
Most of the small drives now will survive a 20G+ force when running
and over 100G when off. Humans quite often die at 20G's. Well
your heirs could have it then.
Scot Harkins at Thurman Industries
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote in message
news:377e46b7...@news.ricochet.net...
>
> Ouch. Disasters are no fun. In the earthquake of 1989, I lost a few
> hard disks because the servers were on a UPS. The UPS kept the drives
> spinning while the ground kept shaking. Eventually, the heads bounced
> off the platter killing the drives. If the UPS had shut off at the first
> sign of a quake, the drives would have survived. There's a moral here,
> but don't ask me what it should be.
The moral is to be prepared for any circumstance. APC (and others) ought to
have a module for detecting earthquakes that will trigger an expedited
shutdown. The trigger could be deactivated for times when the unit
containing the detector must be moved while under power.
Given that earthquakes usually last only in the seconds (up to a minute) the
"expedited shutdown" may just be a cutoff of power. Hard drives take a fair
amount of G's when silent, but not much when in operation, no matter what
the marketing says.
Scot
--
Scot Harkins (KA5KDU) | Systems Administrator, Thurman Ind, Bothell, WA
North Bend, WA | Native Texan firmly planted in Western Washington
sc...@bigfoot.com | SCA: Ld. Scot MacFin, Barony of Madrone, An Tir
sc...@scn.org/msn.com | URL <http://www.bigfoot.com/~scoth>