The picture above is a "portable" air conditioner. This particular one is rated to cool 17000 btu/hr and is pretty much the largest air conditioner you can run off a standard 20A circuit (it pulls 16A when running). It's one of two (the second one is larger) attempting to keep our second server room cool.
While we had planned on just renting this particular unit until the permanent AC system was installed, it became obvious quite quickly that planning and permitting would extend the rental past the point where it made sense to just buy the thing. We now own two of these (the other requires a 30A circuit), and will hopefully never need to rent one again.
That said, anyone that runs a server room should know where to rent one of these, how long it would take to get it delivered, and how it can be paid for. Our local provider (where we both rented and purchased our units) is Atlas Sales and Rentals. Great guys if you need such a thing in and around Silicon Valley.
I was in need of a bunch of specific lengths of power cables to clean up the wiring on one of our racks. I asked my salesman at CDW for "normal PC power cords to go from a typical wall outlet to a typical PC power supply in lengths of 1, 2, 3, and 4 ft".
Apparently that wasn't clear as I needed as I was presented with a quote for not what I was seeking. A little searching (and questioning of my engineer brother), and I learned that I sought NEMA 5-15 plug to IEC c13 cords. And in case you need to figure out the names of a power connector you find on the wall or on a piece of computer equipment, checkout, Wikipedia's pages NEMA and
IEC connectors.
Let's say you order a laptop with a gig of ram. Let's further suggest that you paid extra to have that gig in the form of one stick of 1gb instead of two sticks of 512mb.
Now consider that many laptops have one ram slot under the keyboard and that slot is often annoying to get to, while the other slot is often readily accessible with the removal of a single screw on the bottom of the laptop.
With this knowledge, which ram slot would you expect the ram to be installed in? Can you guess which one Dell puts it in about half the time?
That picture is indeed of a half height AGP video card with the mounting bracket removed and the connector cut off. The surgery was performed to cope with an annoying problem of many x86 systems; that of requiring a video card to be present to boot the system.
I have of course never used this card in a production system.
In the last two months a bunch of users have come to me telling me that there laptops were giving them warnings about their batteries reaching the end of their useful life. When the first two were reported, I assumed they were unusual flukes just gave the users new batteries. When the third came in, I suspected something was up.
With six basically dead batteries in front of me (the oldest only 14 months old) I give Dell technical support a call. Once again I learn that Dell can't track an items serial number back to an original system or order. I also learn that Dell has a one year warranty on batteries. The support rep suggests I check to see if the batteries are part of a recall at the Dell Battery Program.
Nope, none of the batteries are part of a recall. With service tags for some of the batteries, the support rep looks to see if the batteries would still be under warranty. Nope, not under warranty. Requesting his manager gets me the response "There are some recalls involving runtime issues that I can look up". I guess the Dell Battery Program page is only for recalls that might be dangerous to the user.
It turns out to be my lucky day, as the rep can replace all six batteries under one of those apparently top-secret recalls. With a case number and a promise for new batteries to be on my doorstep within two business days, I was off the phone in under 30 minutes.
Sitting in our server room right now is a new 250 lb toy. I spent a few minutes this afternoon contemplating how exactly we were going tolift that thing onto the rails (it doesn't help that I'm not particularly confident that the provided sliding rails will work as advertised).
I brought this problem up to a friend who works in a large datacenter. He told me of the existance of lifts built to ease the process of hefting large servers into a rack. The only such lift that I could find advertised for sale is one from Server Life Corporation (which seems to cost about $8k). IBM makes reference to a support document pdf doc. Not quite designed for racking servers, but R on I's Lift-o-flex line looks to be a reasonable compromise of function to cost (under $1000 for the largest ones).
Not having a large enough serverroom to justify even the Lift-O-Flex (and certainly not being patient enough to attempt to acquire one before using the new toy), I recruited a couple of burly co-workers and it was heaved 10 inches up and two feet back onto the rails without much difficultly.
So in a tale similar to this one, I once again was in possession of both a new laptop and a broken one that needed to be returned to Dell. In the box with the new laptop was a prepaid DHL label to return the broken machine. I needed to put the RMA number and our return address on the label, and call up DHL to arrange a pick-up.
It took me a little while to get around to calling DHL (I wasn't about to give up the broken machine until I had verified the working status of the new one). The day it was finally picked up by DHL was the day the office manager came into my office with a letter from Dell indicating 1) they were grumpy with me for not having returned the broken machine and 2) that if I have already had the machine picked up by DHL that I needed to contact Dell with the tracking number for the package. Once I provide the tracking number (and it's in transit) Dell will stop pestering me.
It would have been nice of Dell to fill out the prepaid shipping label with address and the RMA number, but I can forgive them for that. It is however pretty sad that they can't keep track of the tracking numbers themselves. Apparently Dell doesn't understand the little touches about customer service.
The first report of a problem with the laptop was that the screen went "funky". Our advice was to hibernate the system and power it back on. If it occured again, the user should call us when it occured. The problem of course recurred, and I was called. The drive was moved to a new laptop and the user continued on their merry way. A call to technical support went smoothly, with the result being the laptop being shipped to the repair depot, where they replaced the screen.
Fast forward a month, I am setting up the laptop for a new user and the screen problems return. I deploy a different laptop the new user and make another service call. The laptop takes another trip to the depot, where the motherboard is replaced.
Fast forward a month, the laptop is out on short term loan to an engineer on vacation. He calls me up and tells me the screen is "woogie". He was able to work around the problem for the week by simply suspending and awaking the machine whenever the screen went south. A trip to the depot replaces the screen (again in theory).
The laptop is returned to us in a worse state than it was sent it; some of the damage likely from the poor packing job by the repair depot. My call to technical support reports a broken wireless switch (it falls off), a flakey screen (same problem), and a horrible whine. It only took me 20 minutes to get them to send a technican onsite to make these repairs (as opposed to sending it back to the depot again). I am told that I will be contacted by the repair person the next day and they would be out to do the repairs the following day.
When I haven't heard from the repair person by the end of the second day, I call Dell. Who can't escalate the matter as the people it would be escalated to have gone for the weekend. I am promised a call on Monday morning from the support escalation team. Monday morning passes and I finally hear from the repairman. He asks when would be convenient for him to come out. He doesn't catch the sarcasm when I reply "Three days ago". We setup an appointment for the following day.
The next day arrives, I show the repairman into a conference room, where he opens the boxes with the parts and learns what work he will be performing on the laptop. In about an hour he replaces the screen, motherboard, and wireless switch. I check the laptop out and everything appears to be in working order. About 30 minutes into a Windows install, the screen problem comes back. About 30 minutes later the whine comes back. The trip from the repairman however did yield one useful piece of information; after the second repair, I should have called Customer Service instead of Technical Support.
That's just what I did. They listened to my complaint, they offered to transfer for me to Technical Support, I asked for a new laptop, and 20 minutes later the agreed. Four days later a new laptop shows up.
Too bad the laptop only vaguely resembles our standard issue laptops. The CPU is faster (great), there is more ram (no problem), the hard drive is bigger (sure), it has the Intel wireless card instead of the Dell one (annoying as I like consistency), the Nvidia graphics card instead of the Intel one (seriously annoying for the automated installer we use), has a fingerprint reader (annoying in that it is a driver that I now need to deal with for just one laptop), has a dvd writer instead of a dvd/cd writer (sure, whatever), has the extended battery (eww, it makes the laptop bigger), it has a more expensive version of Office (not cool as it is yet another special case I have to track).
Sigh. I guess this will become my test/development machine.
My boss recieved a phone call from our alarm company about 8pm one evening. He was informed that the temperature sensor in the server room tripped. They didn't know what temperature it was in the room, they only knew that the sensor had tripped and that it was still tripped. When he arrived, the room was 86 degrees. It seems our newly installed AC unit had failed.
As the backup (original) AC unit could not even begin to keep up with our processing machines, he relocated all of those machines to the secondary server room. Thinking the problem temporarily solved, he retired for the night. When I arrived the next morning, that secondary room was almost 80. I arrived about the same time that all of the engineers were arriving and were beginning to use those processing machines more seriously. Within an hour, the room was over 90 and a few of the cases were quite hot to the touch. Once again at a critical temperature, I powered off the machines and investigated my options.
All of our servers are powered by Opteron's, and nearly all of the motherboards support PowerNow, we double-checked that all of our processing machines were configured to take advantage of this dynamically adjusting clock speed (HP's PowerNow for linux instructionsGentoo instructions. With the primary room, maintaining a mid 70s, and the secondary room staying above 80, we forcibly set all of the processors to the slowest speeds available.
Later that day, we had new circuits installed in each of our server rooms to handle our two ton Movincool. And our primary AC unit had been inspected, with a slated time to fix it being the next day once new parts arrive.
Now a portable AC unit like the Movincool is a reasonable backup unit (it had previously been used as the primary cooler in a previous server room), it is not really the best solution. The correct solution is to install two HVAC systems in the server room, either of which could handle the full load of the room. Then, run each of them at half utilization. One day.
Below is a picture of a server (click the picture for a bigger version). This particular server has space for six sata drives, a slimline optical drive, and a floppy drive. Guess where the floppy drive is.
We have nearly one hundred 19 in. Dell flat panel monitors across four revisions of the hardware. A month ago I would have happily recommended the monitors to anyone. I just boxed up the 21st monitor that has been replaced because of screen burn in issues. There were bad monitors across nearly every batch of monitors purchased over a two year span, and across all four hardware revisions. I no longer recommend Dell flat panels.
My first indication of the epidemic occured nearly two months ago when I swapped out a pair of badly burned-in images for a user (which prompted a few other people to complain of the problem). With a half dozen bad monitors sitting in my office I finally got around to calling Dell. I went into the call expecting this to be simple. The monitors have unique serial numbers. I assumed I would spend some time on hold and then give the nice support technician a list of serial numbers. They would then tell me which monitors were still under warranty and which were not, and set up delivery of replacements for the ones that were. If only it were so easy.
Over the course of 4 hours, 3 phone calls, and about 10 different support
technicians I learned a few things. 1) Dell does not repair monitors. 2) Support technicians can not look up warranty information on anything but express service tag (noting that express service tag numbers only come with computers and laptops). 3) Dell does repair monitors. After all of that, I still didn't know if any of the monitors were or were not under warranty.
The next day, my boss gave it a shot. He lasted about 30 minutes before giving up on tech support. In the end, we asked our salesperson to resolve the matter for us. With no response from the salesperson, we rejected the delivery of a fairly sizable order that was being delivered from Dell. We were then put in touch with a very helpful customer service representative who has helped replace all of our bad monitors without hestitation or further wasted time.
It's too bad all of our replacement monitors had been previously loved and half
of them were sent to us in conditions that should never had been it through
Dell's quality assurance group. Nothing serious, just things like damaged
cables, poorly packed for shipping, not including cables. As you might have
guessed by this point, we are no longer purchasing Dell monitors.
We use a good of software that is locked up by FlexLM. FlexLM is a license management and enforcement system sold by Macrovision (formerlly Globetrotter) to makers of software. The system can enforce all sorts of policies; most of the time it either locks a program to only run a specific computer (tied to mac address, hardware dongle, ip address, etc.) or allows a vendor daemon running on a server to provide a certain number of client workstations to run the software concurrently.
Each of the nine application suites that we use that use FlexLM have a license file that is tied to a MAC Address. As part of our efforts to clean and make sane our critical infrastructure we made plans to move the FlexLM daemons to a virtual machine. Since VMware does not by default guarantee that a MAC address for a virtual machine will never change, I followed the best practices laid out by VMware to manually set a MAC.
The short version of that best practices document is that the range 00:50:56:00:00:00-00:50:56:3F:FF:FF is available for assignment by the end user. I choose 00:50:56:00:00:01. It seems one of the vendors of an application thought it was fake and questioned it. Oops, I hadn't thought about that issue. Cutting and pasting the output from the ifconfig command put an end to the complaint.
A few Friday's ago, at 11:48pm I recieved an email from our new Network Appliance filer indicating that a hard disk had failed. The subject was "FILESYSTEM DISK NOT RESPONDING". Shortly there after, I recieved an email from my boss (who was at the time in the process of transitioning to said new filer);
"Score! A disk failure in the middle of the rsync."
A bit later (at 1:35am) we recieved an email from Netapp asking us to confirm the address we wanted the new drive sent to and to confirm that someone would be there for the next several hours. It seems we have four hour repair service for our filer, and that includes getting us replacements on weekends and in middle of the night.
Slight Tangent: The drive traveled less than 10 miles from a UPS logistics warehouse to the company. Had I as a random person paid for that UPS SonicAir service, it would have cost nearly $150. Woweee
Last week, I ordered a server to run a tape jukebox and perform backups
for all our servers. To save a little money, it was ordered without
drives, as we have a small pile of 250 gb drives sitting around after
having upgraded a decent sized sata array.
The server (in one of these chassis)
showed up, I set it in a rack, and went to install a pair of hard drives.
For some reason the drives weren't being recognized. Upon further
investigation, it seems the layout of the bays up-front do not match the
locations of the sata connectors on the back of the back plane.
The bays up front are numbered
0 3
1 4
2 5
The ports at the back of the back plane are numbered
I wanted to add a pair of hard drives to a server. I had the drives, but
I needed a few of Dell's custom mounting trays to use them. My sales
rep sent me a quote for the parts (not available through the website
without a hard drive it seems). The trays would be $10.95 each. There
was another item with on the quote; 8 "SCR,6-32X1/4,FLH,MS,ZPS,CTSK" at
$.05 each. I asked the sales rep about it, and he said they were screws.
I can't believe Dell is going to bother to charge me for 20 cents worth
of screws. Why didn't they just add another buck to the cost of the
drive tray and call it done. Heck, they could charge $20 per tray and I
wouldn't think much of it.
Amusing Tangent 1: The quote came with a from
address "@del.com". I can understand them owning that domain, which they do, it
should be a silent redirect, and it looks quite unprofessional to use it
for email.
Amusing Tangent 2: I placed the order and recieved a confirmation
email a few minutes later. This confirmation showed that I had ordered
2 hard drive carriers, 8 screws, and 5 of my salesrep. I sure hope he
can share with himself as the only space I have for him at work is a
small paper closet.
I have always felt my server room has been hot. Informal observations with a simple thermometer showed temperatures hovering in the mid to high 70s, with not-infrequent forays into the low 80s and the rare spike to nearly 90 on days when the air conditioner stops spitting out cold. With money in this years budget that has not yet been planned for it was time to consider replacing the inadequate window air conditioning unit with something more appropriate. But before I make plans to spend to spend a few thousand dollars on an air conditioner and installation labor, I needed more solid data.
I sought an inexpensive (less than $500) device that could handle at least four temperature sensors, required no server side software, and could be queried by my cacti host (preferably via snmp). As far as I could find, there were two options; the APC
Environmental Monitoring Unit and the IT Watchdogs WeatherGoose. (Note: it seems APC is replacing their own environmental monitoring line with the products of the acquired company NetBotz
I ended up choosing the WeatherGoose (online demo) as it more easily handled more than two remote sensors and it provided a cleaner interface and simpler ways to ge log data out of the device. With a 30 day satisfaction guarantee, I placed an order for the base unit, a door sensor, and two remote temperature sensors, all for a little over $400.
Installation would have been painless, had I not had to fish some of the probes through a suspended ceiling without the appropriatetools. Not including pulling the sensor cables through the ceiling, I was seeing data on the web interface (demo) in under 30 minutes. All in all, my only real complaint is that the unit has a damned wall wart. More on more Cacti setup real soon.
While I am confident in my onsite backup system, to handle the possibility of my server room going up in my smoke, I need offsite backups. I have about 600 gigs used on my primary backup server, but I only need to store about 300 of that at an offsite location to be able to recover from a complete server room meltdown.
Taking tapes offsite is of course the classic way to solve this problem. This means user intervention on a regular basis (which means it is more likely to be skipped), and it means using tape, something with which I have a great dislike (mostly caused by bad experiences with QIC tapes in the early 90s).
I could contract out to an outsider service, and the central IT group on campus will sell me storage on a massively redundant Tivoli managed backup system for two dollars per gig per month. That's $600 a month or $7200 a year for my 300 gigs.
Or, I could build a server and host it some place else. Turns out I wasn't the only person on campus who had a desire to host a server in an "offsite" (meaing not in the same or an adjacent building) location. I easily found another systems administrator on campus who was willing to swap space in our respective server rooms.
For $2500 I can build a server that will meet my expected future needs for at least three years. Sure the server likely wouldn't be as robust as the Tivoli managed service, but I don't really need that level of service, I simply need another layer of protection.
I heard beeping from the room next door. It was not a beeping I heard
before. It almost sounded sickly. Poking my head into the server room I
hear for certain that the beeping was not just an angry piece of
equipment, but was an unhealthy angry one. The beeping fluctuated,
growing quieter and quieter before making a feeble attempt to shout for
attention at its former volume.
It turns out the beeping is from a UPS. Further investigation found
that the UPS had failed in a serious way as the monitor and kvm plugged
into it were no longer getting power. Moving those devices over to the
backup UPS (we are not big enough to justify a proper failover system), I
take note of the serial and model number of the dead UPS.
I am on hold with APC's technical
support line for less than five minutes before speaking with someone.
After hearing my explanation he asks for the serial number and asks if the
battery pack has been replaced in this unit. The tech was quite surprised
to hear that it had not been considering the unit was manufacturered in
late 2000. He went on to explain my options, providing me part numbers
for each; purchase a new battery pack (about $50), get a ChargeUPS pack
for the unit (about $90 and extends the warranty of the entire UPS), or
trade the unit in for a larger model ($50-100 discount off of MSRP). Kudos
to APC for a well run support group.
That phone call went exactly like it should have.
With a new battery in hand I set about pulling out the old one. After
a good deal of effort, I finally get the dead battery pack out, only to
find the top of the battery seriously distended and out of whack.
The new battery went in fine, charged, and went through the basic tests
provided by PowerChute
Personal Edition.
I really dislike wall warts.
Primarily because they take up so much space near plugs. But also
because; they seem prone to fail more frequently than integrated power
supplies (although this is likely because most wall warts are cheap linear
ones), are hot (also likely because of cheap build quality), waste a good
deal of electricity (once again the fault of cheap warts usually), and
easily come unconnected from the device they are powering when wires are
accidentally (or intentionally) jiggled.
Under my desk at home I have 9 wall warts (Palm, camera,
usb hub, battery charger, external hard drive, cordless
phone, dsl router/modem, and two ethernet switches), connected to a
single 7 outlet power strip via Power
Strip Liberators. Even with my nearly obsessive need to tie up excess wire, it is a mess.
At work I deal with a handful of wall warts at my desk for things like
my USB hub and ethernet switch. But it is not these that caused my
displeasure to bubble forth out of my brain. My complaint is about things
marketed toward businesses and things designed to be mounted in racks Even more specifically,
my complaint is about a
Belkin OmniView 8 Port KVM I have.
Now it is my understanding (gleaned mostly from my electrical
engineering brother) that wall warts are used because; companies can use
off the shelf power supplies, the design of the device is easier as they
don't need to deal with interferance or heat from the power supply, and
they don't need to get their device certified by places like the Underwriters Laboratory since it is a low
voltage device.
Wall warts on my server shelves are a pain. UPS's are
not designed to accommodate them. They are more difficult to tie up
neatly (I purchase cables of the correct length so there is little to tie
up normally). And perhaps the biggest gripe, the damned barrel plug on
the KVM falls out at the slightest nudge (well, it did until I applied a
dab of hot glue to the top of it).
As I now look for environmental monitoring hardware I see wall warts
everywhere. Wall warts on $500 products simply do no make sense. So I
make a plea to all electronics makers; please eliminate wall warts
whenever possible. If not possible at least use high quality inline
warts. While I go out of my way to purchase products without warts for
work, I will file a complaint with any product maker that uses them when
the size of the device could accommodate an internal power supply.
So along with 40 laptops we use as a mobile instructional lab, we got
four external battery chargers for the batteries used in those laptops.
I have been using one of them for a few months now. This morning I
dropped four batteries into it. Coming back a few hours later
expecting the green lights to be blinking (meaning the batteries
were charged), I was surprised to see three of the four lights red.
Now I had understood that that indicated the batteries were bad in some
way but didn't quite know what it actually meant.
A little bit of trial and error found that any batteries put into those
slots were now deemed bad even if they had previously charged fine.
At this point I go to the Dell website looking for the manual for the
charger. Searching for battery charger didn't come up with much. Finding
a part number (0F0075), a service tag, and an express service code, I go
back to the support site and try to look up the device by those.
Support.dell.com tells me that the Express Service code and the service
tag are not valid. A general search of the site brings up a page with an
interesting tidbit.
Q: Does Dell sell external battery chargers?
A: Dell does not currently offer external battery chargers.
from Dell
Notebook Battery Center FAQ
Wow, this must be a pretty serious hallucination I am having.
Searching the internet brings up nothing but replacement batteries. So I
call up Dell.
A brief tangent on phone mazes: I have dealt with two phone mazes
today, one at American Airlines and the
other at Dell. Both systems allowed
(and basically required) voice response to the automated questions. The
one at Dell could barely handle exact responses as requested. The one at
American Airlines on the otherhand could handle all sorts of extraneous
noise and responses. For example it could figure out that "yep", "yes",
and "correct" were the same thing. The one at Dell could only handle a
nice short enunciated "yes".
While upgrading their software, Dell
should invest in a decent audio to hold music adapter. The hold music was
pleasant generic light classicalesque, but it faded in and out and was
usually crackly. It would also be good if they normalized the volume of
the music to the lady that continualy informs me that "All of our
representatives are assisting other callers. Please remain on the line and
a representative will be with you as quickly as possible." was a similar
volume to the music.
After asking and confirming my express service code, the phone maze
directs me to Dell Plasma TV support. I quickly get bounced to another
phone queue followed by another, and another. Before I actually speak
with someone that can help, I have spent over 2 hours on the phone and
have spoken with at least 4 people.
During this enjoyable time on hold (cordless phones with speaker phones
are required for dealing with tech support) I discovered another
interesting issue with the battery charger; apparently it needs to be
powercycled. After it was unplugged for a minute all of the charging
slots once again seemed to work and all of the batteries were once again
deemed good
Sure enough, once I finally did get ahold of the manual(pdf),
the troubleshooting section on page 12 suggests powercycling as a solution
to the red light indicators.
Having decided upon m0n0wall for
our firewall, I set it up in the building machine room with a
small form factor Dell and an
Intel
Pro/1000 MT Dual Port NIC. Even before getting grief from the
campus IT group about installing a non rack mount machine into the
network rack, I knew I would need to replace it with a real rackmount
server at some point.
I created a requirements list for the server.
1 or 2U rackmount case
Sliding rack rails would be nice but are not required
4 network connections minimum, preferably 6. No more than two can
be 10/100, the rest must be gigabit.
The gigabit connections should be connected to the motherboard via
pci-x or pci express.
Preference given to Intel NIC's, however Broadcom and 3com are
likely acceptable as well.
The server should be able to boot from cdrom or floppy.
The motherboard must have pata ide on it.
The hard drive should have the cheapest drive that can be
purchased, as it will be replaced a compact flash ide adapter. (like
http://www.pcengines.ch/cflash.htm)
The motherboard and processor should be considered servergrade.
The processor should be at least 2ghz.
512megs of memory is fine. This should be in the form of one
512meg stick of ram, unless the motherboard can make use of dual channel
memory, in which case two 256 meg sticks is acceptable.
The system should cost less than $1500
Knowing that Dell
couldn't really provide what I wanted, and wanting an excuse to look
into some of the whitebox builders out there, I shopped my requirements
list around. In general I recieved quotes in just a day or two. For
every quote, I needed to follow up with a list of questions clarifying
or verifying various items on my requirements list. The contentious
issue was almost always how the gigabit ports were connected to the
processor.
One of the responses surprised me. I had asked for the specific model
number of the motherboard used so I could look up the chipsets used and
verify the bus used to connect the onboard network controllers to the cpu.
Instead of those answers, the salesman apologized and said that his
company could not provide the machine I requested. While this was the
most extreme "wrong" answer I recieved, I wasn't particularly happy with
any of the quotes I recieved. Mostly I found that salespeople don't seem
to like answering questions that are "hard". While I still have a few
months to check more possible whitebox companies, currently it looks like
I will be putting together my own machine based on a SuperMicro
bare bones system.
A week before classes, I did a run through of setting up and using
our new mobile laptop lab. Most of the issues were related to the
logistics of setting up a projector in a classroom not designed for a
projector. There was however a problem that I had a hard time believing
was real.
One of the classes that will be using the laptops will be using them
as dumb terms to connect to a remote x server. Everything was working
as expected until we tried to run SAS.
That generated the following error in the console window;
ERROR: Floating Point Zero Divide.
ERROR: Generic critical error.
ERROR: Explorer failed to initialize.
WARNING: Protected resource may be inconsistent
WARNING: Protected resource may be inconsistent
The campus has a site license to X-Win32 which we use as the
local X server. It provides a properly setup ssh client to connect to
a remote x server. All other X windows applications (xeyes
of course being one of the more important ones tested) displayed
correctly and worked properly. It was only SAS which caused a problem.
Through a good deal of trial and error (testing different computers,
different accounts, different servers), we discover that the error only
occurs when a projector is plugged into the laptop prior to turning on
and logging into the laptop. To further add to the oddness, it is only
an issue when an "intelligent" (which is to say a modern one that does
autodetection and automatic image optimization) is used. If a video
splitter (which will strip any communication between the display device
and the laptop) is put between the laptop and the projector, SAS will
run without error.
SAS's tech support provided the following answer and suggested
solution
"We've seen this problem only with Sharp so far, but I can
tell you what is happening. The video drivers in your projectors are
corrupting the SASUSER profile catalog. It does sound far-fetched, I
know, but it's true.
Some things we've found that work are:
1. Starting the projector first, THEN invoking SAS.
2. Turning down the hardware acceleration on the projector's video card.
"
For the time being, we will be using an older projector which doesn't
cause this error to occur. Bug reports have been filed with both SAS
and Starnet, not that I expect much to come from either report as both
companies will likely blame each other.
Note to those reading via rss/atom, there are images in this post that you likely can't see.
Computer locks are a necesary evil in many places. The computer lab I support is one such place. We do not go so far as to lockdown the keyboards and mice (I have seen places that do this, and it usually makes the working conditions for users much worse), but we do lockdown the computers, monitors, printers, and such.
Right now, we use like keyed normal Master Lock padlocks, and like keyed Master Lock padlock cables (as shown here)
The computer is locked closed with the padlock, which has the cable lock run through it. The cable is looped through the stand of the monitor and locked to the desk. (shown below)
Now those familar with Dell flat panels most likely realize that this is not secure, as the flat panel itself is attached to the stand by screws that would stop a thief for maybe a minute or three. Dell would have us use the thoughtfully supplied Kensington Security Slot (slightly blurry picture below).
Now the lock required to use that security slot (picture of a typical example below) would cost a normal consumer $40. A like keyed set of 25 would cost about $30 a lock. That's quite a bit of a difference than the $2 padlocks and $15 cables we bought in the past.
A slight tangent: I don't actually dislike the Kensington
Security Lock system. I think it is wonderful that there is a sane
non-mandated standard for locking down portable devices. I even carry one of their retractable locks with me in my laptop goodie bag. What I object to, is the price, and the instance on using it for larger devices. A monitor is easily large enough to accommodate a security hole large enough for a standard padlock. It would certainly be trivial to include a decent locking point on a 100 pound server, yet Dell did not in the most recent server we purchased from them.
Unfortunately it does not appear as if we are going to have much choice in the future. The current generation of monitors from Dell have a stand that can be removed at the touch of a button (and thus necessitate use of the Kensington locks), and their desktops use custom lock mechanisms (see below for an example) (that custom lock mechanism from Dell costs $30 btw).
Apparently I will have to factor in a few thousand dollars for locks and related bits and pieces when we next upgrade the lab.
When I arrived in this job the half dozen servers were spread across four tables in a cramped room with a window air conditioning unit. The room was a tangle of wires; the UPS that was physically closest to a server was not necessarily the one it was plugged into. There were ethernet cables being strung in from other offices via the hung ceiling.
Doing something about the room was immediately put onto my list of big projects. While I am pretty sure every good admin dreams of nicely rackmounting everything, it can be an expensive route to take. With a good rack running $800 and rack mount equipment for servers costing $100-$200 per server, the costs add up quickly. The per server cost is reduced if the rack mounting hardware is purchased with the server. Racks are unfortunately not really an option for me however.
The initial cost (which I estimated at $2200) was only a small part of my decision to not rackmount everything. My biggest reasons for going with strong shelving instead of a rack was the need for flexibility and the lack of need for density. Racks make packing many servers into a very small space easy. I have plenty of space and no need for more than about 10 servers at any one time. Flexibility was really the deal breaker for me.
We had (and still have) a strange collection of servers and related hardware. With good shelves I don't need to worry about how or where I will store something when I consider a purchase.
It has taken a good deal of patience and nearly a year, but the room is
almost organized.
The
shelving is Metro Super
Erecta. It's 30 inches deep, 74 inches high, and 60 inches wide. In
theory each shelf can hold 600 pounds. That does actually hold pretty
true in the real world. The heaviest loads should be closest to the
poles to avoid sagging in the shelves. Also important to note, is that
this is the commercial version of the Metro shelving and not the consumer
stuff you may have seen at your local mega-homestore.
The back
of the eight port KVM with neatly tied down cables. Nearly every wire on
the shelf is tied down with cable ties (yellow ones seen in this picture)
and labeled at both ends (white tags visible in this picture). The
labels were made with a Brother Ptouch 2600.
There is an
admin console and a KVM switch to access all of the servers and computers
on the shelf. The KVM is an 8 port Belkin (nothing designed for
rack mounting should use a wall wart).
There are five
UPSes on the middle shelf, four 1500A APC
SmartUPS and one 750VA one. We only need two of the four
1500VA UPSes to run all of the servers off of batteries for around 30
minutes. Having four means we can run off of batteries longer and it
means that we can have up to two fail without loss of protection (this
does assume as in our case that all of your servers have dual power
supplies and they are plugged into different UPSes). The 750VA UPS is
used to power the console and the in-room ethernet switch.
There
are still some wires to tie up and neaten. That should be easy to fix,
as soon I find a source for various lengths of decent power cords. The
window air conditioner is another matter all together. That is a
subject that shall be saved for another time.
I just bought a Dell Server. I have had my eyes on a server for over six months now. It is going to be the other half of a pair of VMWare GSX Server hosts. It will provide a means for reasonable disaster recovery, a good deal of room for expansion, and the means with which to eliminate at least one physical server.
I originally spec'd the machine with dual 3.6Ghz Xeon's, 6gb of ram, four 146gb 15k drives, two 73gb 15k drives, and 3 years of gold service. The Dell website told me it would cost $11,000. That price was reached through the Wisconsin State Dell Store. The same machine configured through the Small Business store put the price at $12,500. The Medium and Large Business store didn't give me the options to configure a machine with the same specs. We ended up paying less then $8,000 after speaking with the Wisconsin Dell representative.
In a way, I like buying from Dell. It's always a pleasant surprise to spec a machine at the price I want to buy it at and have it cost much less. But it is also an exercise in abusive selling tactics. I really dislike it when companies screw with their customers.
I would love to be rebellious and say that I have no intention of purchasing from Dell again (this time I had no choice as I had to have a server that matched certain specs of my other VMWare Server host), but it is not quite that simple when purchasing for a university. There is political pressure to purchase from University approved sources. There is an economic advantage (Dell provides us excellent rates) of purchasing from the State's primary computer supplier. And finally, the State contract provides warranty and service from Dell that simply can not be had for a similar price.
One day when I actually put out most of the fires that are raging, I will have the time to persue purchasing from other vendors, perhaps whitebox ones where I get to choose the specific components.
So every Dell produced since at least 1998 has had a service tag (which I faintly recall being called express service codes back in 98), a 5-7 letter and number code that uniquely identifed the machine. The code is printed on a sticker somewhere on the outside of the case. Newer machines have a sticker on the side which includes the service tag, date machine was produced, and other such useful tidbits.
Now with a unique identifier one would think that it would be easy to lookup the exact specs of ones system as it was purchased. Dell's support site does indeed let you do this. Sort of, if you can figure out what lines like this mean...
1 K3444 CARD (CIRCUIT), NETWORK, MINI PCI CARD, INTEL2200, NOT APPLICABLE
In this instance it means an Intel 2200 minipci wireless card. Usually dechipering these descriptions isn't too hard. Here is an example spec sheet for a full desktop system.
Service Tag: 7KYPQ01
System Type: OptiPlex GX400
Ship Date: 7/19/2001
Dell IBU: Americas
1 43YVH PROCESSOR, 80528, 1.4GHZ, 0K, 400FSB, SOCKET W
1 57589 CABLE, AUDIO, MOLEX TO MOLEX
1 6F067 PRINTED WIRING ASSY, PLANAR (MOTHERBOARD), ASTRO, NETWORK INTERFACE CARD/CONTROLLERS, 4RIMM, OPPLX
1 9019C ASSEMBLY, CABLE, DORADO/ATHENS/TUALATIN/ALMODOR, 34P, FLOPPY DRIVE, 1DROP, HUNNICUT/MEDIUM DESKTOP
1 9809T ASSEMBLY, CABLE, 40P, IDE (INTEGRATED DRIVE ELECTRONICS), 2DROP, LS120 FLOPPY DRIVE
1 5120P CORD, POWER, 125V, 6FT, SPT2, UNSHIELDED
1 25PGG KEYBOARD, 104, 6P, UNITED STATES, NMB, MIDNIGHT GRAY
1 735HE MOUSE, PERSONAL SYSTEM 2, 6P, 2BTN, WHEEL, 1.3A, MICROSOFT, MIDNIGHT GRAY
2 1561P RAMBUS INLINE MEMORY MODULE, 256, 400M, 128X18, ERROR CORRECTION CODE, 16C
2 9578D CARD (CIRCUIT), MEMORY BOARD, MEMORY, PRINTED WIRING BOARD, CONTINUITY, RAMBUS
1 70NYT COMPACT DISK DRIVE, 128K, I, 5.25" FORM FACTOR, 48X, LENGTH/LONG, 8482B, MIDNIGHT GRAY
1 06HRM DISPLAY, FLAT, D-INTFC, 17, 1701FP, UNITED STATES, MIDNIGHT GRAY
1 34MCW CARD (CIRCUIT), GRAPHICS, VIDEO, 16MB, TNT2, PROFESSIONAL
1 5828D ASSEMBLY, CABLE, ATA66, 2DROP, KLINGER
1 903DP HARD DRIVE, 40GB, I, 7.2K, 20/P, IBM-ERC
1 4C496 FLOPPY DRIVE, 1.44M, 3.5" FORM FACTOR, 3MD, NO BEZEL, NEC CORPORATION, MIDNIGHT GRAY, 418
1 0C138 REMOVABLE MEDIA STORAGE, ZIP DRIVE, 250M, I, 3.5" FORM FACTOR, IOMEGA, V4, MIDNIGHT GRAY
1 97MNT KIT, COMPACT DISKETTE, OPERATING SYSTEM, WINDOWS 98 SECOND EDITION, ENGLAND/ENGLISH, A01
If you tell the Dell support site a service tag, your experience
will be customized to what is appropriate for that machine. Well,
that's my dream anyway. It is tailored to that machine type, not the
specific machine. This is an important distinction, as when looking at
the downloads available for the machine type you are shown all the
possible downloads for all the combinations of standard hardware the
machine sold with. This is annoying when presented with a half
dozen different wireless card choices for a laptop. You can usually
work out what part you have from looking at the somewhat cryptic
system spec sheet.
Sometimes it isn't easy though. I have a laptop, that I know has an
Nvidia video card in it. Dell presents three choices for me to
download. And from past experience, installing the wrong driver will
create graphical weirdness that is only curable with a windows
reinstall. A look at the system spec list doesn't help that much.
My choices for files to download are for an nVidia GeForce FX G05650,
GeForce FX Go5200, or a GeForce4 4200 Go. From going through this once
before (and needing to reimage the machine between mistakes), I know
that it is the GeForce4 4200 Go.
To add to this silliness, the download page will frequently list a
LONG list of optional perhipherals for the machine. For example, the
downloads page for the desktop listed above shows 26 different monitors.
Would it really have been so hard to give me a list of just the
applicable software for me?
