HDD in RAID 10 overheating

I have 4 WD1002FAEX 1TB drives in a RAID 10 array in the hot swap bay on my 800D chassis.

Today one of the drives went off-line. Just disappeared out of the array, showed as "not available" in Intel RST v9.6.

Shut down the system and pulled the drives. The bottom drive was extremely hot, almost too hot to touch. The other three drives were very warm but not what I would call extremely hot.

Marked difference in temperature between the three drives in the top of the stack and the one at the very bottom.

Let the drive cool off for about 2 minutes and put all four back in using the same slots. System came up, all four drives reappeared in Intel RST with one of them now in "rebuild" status.

Suspect what happened was one of the drives got too hot and went into a thermal shut down, took itself off line until I powered off the system and let it cool.

So my questions:

1.) Isn't the cooling fan in the 800D chassis next to the hot swap bay supposed to prevent this? Is there anything I can do to help it out? Add another fan or replace it with one that moves more air? Perhaps a second fan pointed up at the bottom drive?.

2.) Why is it the bottom drive on the stack is overheating? Heat rises so I would expect it would be the top-most drive in the stack that would be having problems.

3.) In a RAID 10 array, is it possible to change the positions of the drives in the array without having to re-dimension the array (and lose all data)? If so, I would like to try swapping the one on the bottom for the one in the next to top-most position, just to see what happens.

Thanks for any feedback or suggestions.

Your drives do not have TLER (Time Limited Error Recovery) which is basically a timeout for a harddrive to correct an error it encounters so that the RAID controller does not drop it. The reason it was hot is because it was trying to correct and error, but for too long (moving data all over the place). Then the RAID controller marked the drive as bad, and dropped it from the array. Its not about temperatures but the actual hard drives.

1. Irrelevant

2. It doesn't have to be specifically the bottom drive, it really can be any drive in your array.

3. Wouldn't make a difference.

Your drives do not have TLER (Time Limited Error Recovery) which is basically a timeout for a harddrive to correct an error it encounters so that the RAID controller does not drop it. The reason it was hot is because it was trying to correct and error, but for too long (moving data all over the place). Then the RAID controller marked the drive as bad, and dropped it from the array. Its not about temperatures but the actual hard drives.

Pengy, can you help a noob at this stuff? After reading the OP I checked my drives and they were clearly the hottest thing in my case by a lot. The bottom one in swap bays was in fact the hottest. I noticed that the fan that is pushing into the swap bays moves very little air, so I put it on an unthrottled power feed. Moves a little more but not much. The air also has nowhere to go. The compartment vents to the cluttered space that hides all the wires. Not good for getting air across the drives.

Here is the question, since I have paid so much attention to cooling my processor and ignored my drives .....

How much heat can a drive take? Rendering video, I have toasted drives before. I hit them hard rendering. Do they need to be kept as cool as a processor or are they more heat tolerant?

Thanks, I appreciate finding folks who know something.

Wrong thread, but do you know if my memory in CPU-Z is reporting a DRAM Freq of 668.3 MHz, does that mean it is running at the 1330 that I was trying to get it to run at or am I only half way. ( A very noob question) :dunno:

http://valid.canardpc.com/cache/banner/1259012.png

With hard drives, you usually want to stay under 60C as a general rule of thumb. Some drives like the black may be able to sustain a higher operating temp, however I do not know the exact temperature. Check the temps of your HD with speedfan. As I do not personally own a 800D (I do want one however) I couldn't say how the hd fan is set up. Actually, encoding does not really use the hard drive that much, most of it is done in the cpu/ram. If you actually check the I/O from your drives, you will find that it reads/writes maybe 6-10mb every couple of seconds because windows cache's the reads and writes to the drive. (using rainmeter and a harddrive graph)

Edit: Those drives were probably heating up because of ambient temps from your cpu/ram when you encode.

Next up is, you need to enable XMP Memory Profiles in bios, or edit your memory speed, timings, and voltage manually. Set your speed to 1600mhz, check in CPU-Z in the SPD tab for memory timings and voltage. Voltage will probably be 1.65, but your motherboard won't allow it to be set there, so set it at 1.66

I appreciate the info. I am checking the HD's with speedfan right now but they look like they are OK temp wise.

Actually I have cooked drives before not due to external temps. I use DMA for the drives rather than cached. I have hit a sustained 90% use cycle on drives and kept that up for over 100 hours straight. Consumer drives just can't take it most of the time. I now spread my material over multiple drives and render to a lone drive. Often I will set up RAID 0's just to keep up and keep it cooler.

I have just finished a torture test on the drives in the swap bay and I'll be ...

There must be enough airflow going through there because they maintained a pretty constant temp with only a 6C increase under the extreme load. All drives stayed well below 50C. No complaints there.

Thanks for pointing me in the right direction.

Yes, the TLER issue. When I sourced the parts for this PC back in May, anticipating that I would want to run RAID was the one thing I missed and as such, bought Caviar Blacks instead of RE3's (at a difference of a mere $30.00/drive).

It is a long shot, but I am going to take a look at the TLER settings on the drives using the amazingly hard to find WD DOS utility and see if they are in fact modifyable. Going through the headache of creating a bootable DOS USB stick now (funny how the lack of an ancient 1.44MB storage format can STILL leave you dead in the water these days).

Good news is, a friend at work will by the 4 Blacks if I can't modify TLER in such a way as to permanently resolve the issue (I take a hit of about $200.00 on what I paid for them). Penalty for not fully researching RAID requirements on my new system :(

(BTW, I do think the hot swap bay in the 800D could be opened up more to allow for better airflow.)

At Newegg the WD Caviar Blacks get very mixed reviews for various RAID setups.

While the reviews at Newegg for the Samsung Spinpoint F3's (1TB and 500GB) seem to get rave reviews for RAID setups.

:confused:

The Spinpoint f3 drives can be enbaled for CCTL (Samsung/Hitachi's version of TLER) and support it. They also perform better than the comparable WD black (FAALS) in RAID and non-RAID setups.

Caviar blacks (Older versions) have TLER disabled on them if they were manufactured in Nov 2009 or later. All new drives like the 6GB/S versions, and WD Green drives have TLER Permanently disabled. Newegg probably still stocks pre-Nov drives in their stock, so they might be selling those, however its all based on luck. Popular drives such as the 640gb and 1TB drives are definately newer drives, 500GB maybe not.

Good to know. You know your stuff. :cool:

I'll be in the market for four single patter 500GB HD's.

I want to replace two 10K Raptors (SATA I, 150GB/sec max) with two Spinpoint F3's in a RAID 0, with a small partition (25-50 GB) for the swapfile and other stuff that would normally be written to the two Corsair P128's which are also RAID 0 (boot drives), the rest of the drive will, of course, have a much larger partition, I doubt that I'll use it much though, but it's there in a pinch.

I'm also thinking of two Seagate drives as primary storage, no RAID, the slim ones (~0.8" thick vs ~1" thick), to slightly improve airflow through the case (the Raptors run a bit hot too).

These drives run ~$50/es, so not that much of an investment.

Kind of off topic I know.

Oh well.