Promise SX6000 Raid Issues

[Bookmage]

So last fall, I decided to get a raid5 hardware controller to create a TB storage array/server. At the time, I was runnin my main server of 3 x 200GB WDJB on an abit vp6 runnin dual vp6. After much research and lookin aroung, I picked up a Promise SX6000 with teh plan to upgrade to 6 x 200GB on RAID5 for a TB of net storage with the possibility of one lost drive. I didn't have the money to shell out for a 3ware 8port and didn't see much point in getting overpriced 250GB drives. Several weeks ago, I picked up several more 200GB drives and started to build the array.

After many problems with the vp6 and BSOD's and the like, I decided to just pick up a new mobo and split the p3s into two storage arrays. After a fresh install of 2K, I finally got the Array setup and formated and initialised. Woohoo 930GB!! However, after 2 days, something went wrong and the card beeped like mad. Upon rebooting and inspection, I was shocked to find drives 1 and 2 had disappeared off the array. I double checked the cables and unplugged them and plugged them back in. Then after I loaded it back up, the Promise BIOS told me drives 1 and 2 were free, and the array offline. I had no idea how to fix this. Since I had nothing on it, I rebuilt it and formated and recreated. It ran fine for about a week, before the same thing happened again. This time, it was drives 3-6 that had dissappeared and are now free and not attached to the array. I have no idea how to fix this. I talked to someone who also had this and they said they had problems with raid5 on their SX6000.
So, is this a known bug? problem? common issue?
Is this with the Promise Card or the WD2000JB drives?
Is it cuz I'm running 4 drives that are new and 2 that are older? diff firmware?

And if this is a known feature, what should I do to fix/solve this dilemma?
Sell the card and buy a 3ware which has dropped in price? Split the drives to two computers each with 4 x 200GB and back them up?

I have 8 Drives available. 4 purchased in the past month, 4 that are over a year old. I have two ABIT TS-20 P3 mobos with One PCI running as domain controllers. I also have an old 3ware 5800 with plans to run 8 x 120GB Samsungs on it for a separate backupo storage server. (cuz Merc says so) The drives sit in some 3 - 2bay fan coolers in a 4u rackmount.

What happens if I JBOD 6 drives and one fails? Do I lose all the data on the array? or just the data on that drive? If I have Files 1,2,3 on Drives A,B,C in JBOD, and drive C dies with File 3 on it, Can I only access File 1 and 2? Or do I lose it all? Is it better to run 2 identical computers with 400GB mirror? or 2 identical computers with 800GB JBOD? Should I just run the 6 x 200GB in JBOD and back up the majority to the 8 x 120GB?
Sell the Promise card and buy the 3ware?


The end result is to have something near a TB of storage for music, anime, and video that is also backed up in case of drive failure. Most will also be backed up on DVDs in due time, but only when it is completed or ready to be finalised. This also serves as a media server for sharing and playing on clients hooked to tvs. Budget is a limited factor and I am unsure how to proceed.

 

[P5-133XL]

We will start out with the easy question. If you are running raid5 and a single drive failes then everything stays intact -- You lose nothing. Everything runs fine even with one drive missing however, if you lose a second drive then everything will be gone. Replace the single failed drive, rebuild the array and you will be back to redundant heaven.

Now why you are having drive failure problems is a unanswerable question at this point. Have you run through Western Digital's diagnostics with the failed drives? you need to check things like cables lengths. Are they all within the 18" Ultra ATA spec? Is you power supply sufficient to power all the drives during bootup (When the power drain is the highest). Are you using molex cable splitters? They have a strong tendancy to supply intermittant power. When everything is up and running vigorously wiggle all the power wires and see if any of the drives shut down. Don't ever forget to check the event logs to see if there are any drive related problems showing up.

 

[Bookmage]

Drives are fine, they have been checked several times. All cables were the bundled cables that came with the card and are 18inches single device. Power supply is sufficient, 550ishw and yes I am using molex cable splitters. The PSU comes with a total of 8 molex 4 pins on two cables, and I have 2 splitters running on one cable. Event logs just show the array going offline and no other problems. It's like the drives just dropped out of the array and disappeared. When they came back, they were unrecognizable by the card. I remember hearing/reading about this somewhere but I don't remember where, or with what hardware, ie card, drives, etc.

Current specs are
ABIT TS20 matx P3 mobo
P3 733 @ 133 under a FOP38 HSF i believe...
512MB Corsair PC133
everything onboard
Promise SX6000
6 x 200GB WD2000JB

 

[P5-133XL]

Whenever a drive drops out, even for an instant, it will permanently be removed from the array. This is normal because it has failed to keep-up with the data transfer of the array and is now unsynced. It won't get re-synced untill you build the array.

From the descriptions it would seem to either be a PS problem or the molex connectors. Try wiggling wires to see if you can get a drive to power-down. Also install a motherboard monitor program that can monitor the different voltages and alarm if they drop from their expected values.

 

[Mercutio]

An oscilliscope and a voltmeter are very useful for this kind of thing.
Anyway, I think in this case you need to be looking at your PSU instead of the drives or the RAID card.

 

[Buck]

What happens when just the newer WD drives are used?

Older WD drives had problems in RAID arrays. They run an internal diagnostic for lengthy periods of time, which causes them to fall out of RAID arrays controlled by a controller card. They still do this (which is good for regular desktop use), but newer firmware has made the drive more available when probed for life my a RAID card. The usual ATA command set would still work, so reads and writes were never hindered. Raptor drives never had this problem. Very early WD drives (prior to 40-gigabytes per platter) did not have this problem.

 

[Bookmage]

Drives 1 and 2 are older drives. Drives 4-6 are new. AT first I had problems with drives 1 and 2 which resulted in drive 2 dying and getting RMAd. Then I replaced that drive with another drive I had from around the same time. That did not have any problems. Recently I added 2 T shaped molex power extensions that fit 3 devices. Each is hooked to three drives that are placed in a 3 to 2 bay hard drive adapter with fan. All plugs have been double checked and pushed firmly. I will test voltages later and see if that is a problem. I originally had an Enermax 431w PSU in there, but that got blown in a rare power outage. So I replaced it with a 600w Raidmax? I think. I'm lookin at replacing that with an Antec TruPower 550. I have created a new array and will test it before I continue to the next stage. A month later, and I'm still testing.... sigh....

What happens if I JBOD 6 drives and one fails? Do I lose all the data on the array? or just the data on that drive? What is the benefit of JBODing 6 drives vs jsut 6 independant drives?

 

[Howell]

What happens if I JBOD 6 drives and one fails? Do I lose all the data on the array? or just the data on that drive? What is the benefit of JBODing 6 drives vs jsut 6 independant drives?

JBOD looks like one drive. So the answer is file/folder/disk management. Today, This is more of a problem in enterprise where data storage rates can increase faster than maintenance intervals would allow for storage management.

Originally it was because you could not get drives as big as JBOD would allow.

 

[Pradeep]

We have a few SX6000's at work. Running 6 * 250GB Maxtors in RAID 5. We had major problems trying to run two of the cards on one mobo, with one array on each card, accessing one array singly was fine. However as soon as you tried to access both, the server went tits-up. But no problems with the drives being detected or dropping out.

Solution was to have one card per server. Not ideal but will do till we switch to some kind of SATA 3Ware or something.

 

[CougTek]

An oscilliscope and a voltmeter are very useful for this kind of thing.

Yeah, but last time I wanted to buy an oscillo, it cost around 1000$ for a middle range unit. A voltmeter is way cheaper, but it doesn't give you anywhere near as much info when you're troubleshooting.

 

[Buck]

Here is what I was talking about regarding WD drives, and now they have a fix -- RAID Drives

 

[timwhit]

An oscilliscope and a voltmeter are very useful for this kind of thing.

Yeah, but last time I wanted to buy an oscillo, it cost around 1000$ for a middle range unit. A voltmeter is way cheaper, but it doesn't give you anywhere near as much info when you're troubleshooting.

I have seen used units for less then $50 before. They aren't top of the line or anything, but it should work for testing computer equipment. Check eBay.

 

[Mercutio]

Specifically, what you want is one of those big, heavy mothers HP made in about 1960. If I was an electronics guy instead of a computer guy, I can't see how I'd ever let one out of my grasp. Built not like a tank, but a battleship - like early HP laserjet printers, HP made them so well no one ever wanted/needed to buy a newer one.

 

[P5-133XL]

Perhaps I'm wrong but do Those old Oscilliscopes have the freq. range needed for modern computers?

P.S. I like the old Textronix scopes too.

 

[Mercutio]

To troubleshoot power problems? Yes.

 

[Bookmage]

Woohoo! drives aimed for RAID and somewhat within budget. But what's up with that 3 year warranty? Oh well, it's a start. I've been running my array in JBOD for a while and so far no problems. I plan on emptying it and changing it back to RAID 5 soon. So we'll see if problems happen again.