ARCHIVES
August 2011
July 2009
May 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
August 2008
July 2008
June 2008
May 2008
March 2008
February 2008
January 2008
November 2007
CONTACT
|
| About this blog: Computers hate me. They really do. Every time I try to do something unusual like add new hardware, something is guaranteed to go wrong. I decided to start writing about my constant problems so that someone else might benefit from my experiences - or at least laugh at them! |
I seem to be jinxed this month - yet another WD drive has failed. This is the third failure in 3 weeks and brings the WD total to 5.
This time it's actually the replacement for the second-last failure, in other words a brand new drive. It's been powered on for less than a week.
This drive "disappeared" (similar to the last failure) at 10am this morning, then when rebuilding the RAID array smartmontools detected the drive had an unrecoverable sector. By itself, a single unrecoverable sector is not necessarily something to worry about, but since the self test clearly fails I'll be returning it.
Incidentally, I didn't do my usual round of tests since this drive was required sooner than expected; it's possible this bad sector was there even before it was installed. Unfortunately the drive that I've used to replace it has also not had any infant mortality testing, since I only received it a few days ago.
At this point I'm having trouble believing that 5 failed drives out of ~20 installed Western Digital drives is an acceptable, or normal, failure rate. Assuming it's not just a bad batch (which is unlikely, since they were purchased at different times) it can only be...
- Physical mishandling somewhere in the distribution chain
- Physical mishandling at retail outlet (they are OEM drives in anti-static bags - no other packaging...)
- UPS
- Power supply
- Vibration (the case has 8 hard drives)
All of these are common failure points since all of the failed WD drives were purchased from the same retailer, and they're installed into the same machine; diagnosing and fixing this issue will be incredibly difficult.
I'm going to pull some more hair out now........
--------------
UPDATE: I believe these HD glitches may have been caused by a faulty "Y" SATA power adapter, which is a single 4 pin molex connector to 2 SATA power connectors. Both the WD 750GB and VelociRaptor "disappeared" together each time, and I suddenly realised why - they were both hanging off the same Y cable. I've replaced the power supply with a model that has 8 SATA power connectors (rather than 6) so all drives are now connected directly to the PSU wiring loom. No problems since. I don't think I'll return these two most recent "failures" as it's likely the power glitch was the cause of the reported bad sector, rather than a HD problem. A zero fill fixed the read failure status, although offline uncorrectable is still showing as 1... smartd reminds me of this every half hour. :)
So I guess we'll disregard the last two failures... which leaves us with 3 "real" WD failures.
|
|
I've submerged a thermometer into the bong cooler reservoir, which means I can now measure its effectiveness. The thermometer cost $16 off ebay so I doubt it's that accurate, but even if its reported temp varies by a couple of degrees above or below the true temperature it proves that the bong cooler is able to chill the water in the reservoir (slightly). Current ambient temp is 32.5C (it's a hot summer night), reported water temp is 25.3C. If I add a reasonable heat load (eg a Q6600 CPU with all 4 cores working flat out) the reservoir water temperature wanders up by about 1.0C; in this case the reported CPU temperature is about 32C, although I'm not sure how accurate that is.
The probe is sitting right at the top of the water level since the cord is a little too short; eventually I hope to have one sensor at the bottom and another near the top for comparison. The pump is submersible so it draws water from the bottom (an area which is hopefully a bit cooler).
I've also made a few tweaks in the past few days and the cooler has become significantly more thirsty - requiring more frequent topups - which is a good sign that it's working.
One thing I'm concerned about is winter temperatures pulling the outdoor reservoir temperature too low. If it falls to single digits there may be condensation issues with cold water being piped into a hot PC...
UPDATE: Daytime temps... ambient 38.4C, water 27.4C (11.0C difference!), CPU 26C (idle)
|
|
Yep, another drive has gone. It started playing up a couple of weeks ago, "disappearing" from the OS a few times as if it wasn't connected. A reboot was required to recognise it again. This time when it misbehaved the "Offline Uncorrectable" and "Current Pending Sector" SMART values increased (ie went non-zero) on boot, and both a short and long self test failed prematurely.
The spare I'd received as a swap for the previous failed drive - replaced around 3 weeks ago - barely had a chance to sit on the shelf before being pressed into service!
It's been about a year since I switched to WD so with further failures there's a greater chance that I will have to return the drive to WD rather than doing a straight swap at the retailer. The catch is that a drive from WD is likely to be "refurbished." I haven't used a refurb WD before, but the Seagate ones are super dodgy: it's almost not worth returning it since they're likely to start playing up within half a year......
|
|
SFU will refuse to install on XP home, but there's a simple change you can make to fool it. Load up your favourite hex editor (search for "hex editor" if you don't have one), then edit SfuSetup.msi and replace the string "NOT (VersionNT = 501 AND MsiNTSuitePersonal)" with "NOT (VersionNT = 510 AND MsiNTSuitePersonal)" (minus the quotes). In other words, change the string 501 to 510 at that location.
I've been using SFU for several months as an NFS client and it works reasonably well. One time it's guaranteed to start playing up is if I reboot the NFS server; NFS clients are supposed to be able to recover gracefully from such an instance since NFS is stateless, but in this case XP will start pausing for long time periods (a minute or more) any time I want to open a file, view a folder etc. No data is lost, but in order to make the computer useable it has to be rebooted. It's annoying, but rebooting the NFS server is an extremely rare event.
|
|
One of the reasons I used a reasonably large reservoir for version 2 of the bong cooler setup was in the hope that the extra water and surface area would assist in passive sinking of the heat, beyond the evaporative cooling process.
I was thinking earlier today that if your room was in the right place you could create a water feature and use that as your sink pond AND evaporative cooler... of course in Australia we're in a drought at the moment which means that water features aren't the most popular things. (It's probably illegal to top them up in most areas)
Here's the large reservoir idea taken to extremes - Water Cooling Computers With A Swimming Pool
|
|
Copper coils attached to PSU. | 
Water piping. |
I've created a crude heat catcher (a radiator in reverse?) using 1/4" copper tubing attached to the power supply fan outlet. Air is drawn through the copper coil by another fan. Water is piped through the coil where it is cooled by a bong cooler located outside and recirculated back in.
I don't have anything to measure temperatures but so far it does seem to be working: there's less "rising" heat around the PSU area (which could also just be the additional fan dispersing it better), and the intake pipe feels cooler than the outlet.
The only problem is the pump: it's pushing through the water very slowly, compared to when I was just using a simple 1/2" copper loop. Head height doesn't seem to make much difference so presumably it's a pressure issue: water moves more slowly and painfully through approximately 4 metres of 1/4" tubing versus 15 metres of 1/2" tubing... there's not much more than a dribble coming out of the top fountain head, rather than a shower. I'm considering using two separate fluid loops: one for the evaporative portion (circulating water from the reservoir, up, then showering down) and one for the indoors cooling loop. The latter can be a sealed system which should be cleaner as it can use distilled water and/or suitable additives. The former loop still needs to use tap water since it will need frequent topping up.
|
|
Been a while, so one was surely due. A WD 750GB drive has half died, with the BIOS reporting a drive error at boot and the OS not seeing it at all. I can't run any diagnostic tools since it's not detected.
I had a spare drive on hand in preparation for such an event, so the RAID1 array is currently rebuilding.
This brings the count of failed WD drives to 3, out of a total of about 20 installed. Still not as bad as the overall Seagate failure rate, but it's not looking so great...
Incidentally, I found the invoice and it seems this drive was purchased at the same time as the previous WD drive failure, which developed bad sectors in June 2008. Coincidence, or is someone in the supply chain throwing these things around? I treat them with kid gloves once I get them. :)
|
|
Outside the office. | 
Water topup solenoid. | 
Water falls to the bottom. | 
Air inlet covered by mozzie filter. |
Here's V2 of the bong cooler. This one has a larger 120mm fan, with the air inlet at the bottom and the fan located at the top (which is now blowing air out rather than sucking it in). The reservoir is also significantly larger, probably around 20 litres at the chosen fill level. A solenoid connected to a garden hose allows remote control of topping up the reservoir. Brass rods are used as electronic water level sensors in order to cut off everything if the water level falls dangerously low (if the whole thing tips over, or someone unplugs the garden hose), and also keep the water at a particular level by filling with fresh water. The micro spray jet has been replaced by a fountain head. Plastic based filter material has been cut into small cubes and placed at the bottom of the tower to reduce the "continous piddle noise" effect - as a bonus it increases surface area which should increase evaporation.
Build quality is also a little better as I expect this to be a semi-permanent installation.
There is still the occasional blip of water leaking from the join just below the bottom elbow, but this is not an issue since it travels down the outside of the pipe into the reservoir. In the V1 design the water would end up running over the lid and off the side.
|
|
|
