Problem: Intel ICH8R raid 5

K

kandor

Gawd
Joined
Aug 24, 2005
Messages
632
Hi,

I have an asus P5B deluxe wifi ap motherboard. It has an intel ICH8R chip on it with 6 sata ports that supports raid 0,1,5 and 10. I hooked up 6 250 gb WD2500KS drives to it and created 2 arrays. Port 0 and 1 are a mirrored array of 250gb. Ports 2 thru 5 are a raid5 array of appro 740gb.

The system has been running fine since I built about 10 days ago, when 2 days ago it booted up and the 'intel matrix storage manager' said a drive had failed. It was the drive on port2. I tested the drive by plugging it into a free sata port on the other raid controller (the mobo has 2) and I could see the drive as unformatted in winxp disk management snapin. It seemed fine so I plugged it back in, booted up and set the drive status to 'normal' in the array manager. It said ok and proceeded to rebuild the drive. Fine.

Now 2 days later the same thing has happened to the same drive. I rebooted, and at first the bios manager showed all 4 drives as 'disconnected' - then on another reboot it showed 3 drives as fine and the fourth drive fails to show up now (the one on port2).

I'm starting to not trust this controller at all - is it just me or is it complete and utter garbage, and why would they put it on a mobo if it's so unreliable?

Any opinions?

Thanks,

Kandor
 
That's the WD TLER bug. Exchange for RE drives if you can; any controller is likely to have the exact same problem.

 
Thanks for the info on TLER - I've read up on it. I can't return these drives so I'm stuck with them. It sounds like I may be better off using the 4 drives for a raid10?

So, to be clear, it's the lack of TLER on my WD2500KS drives?
Also, somewhere I read that WD has a tool for toggling the TLER function but that would only work I'm assuming on the RE versions.

It's happened on the same drive twice in 3 days. I think I'll rma that drive, or is there something else I can do in the meantime, like defragment or scandisk. (Is it sensible to defragment raid5?)

Last thing: Intel's manual for ICH8R says it supports raid 5 with '3 or 4' OR '5 or 6' drives, but when I built the array and tried to select 5 it has a message saying 'select only 3 or 4 drives' - seems like something doesn't work the way it should?

Kandor
 
Raid 10 isn't any more allowable than raid 5. You may be best off running seperate disks and storing .par2 files on one of them :eek: Manual raid5 for the lose...

Yes, the lack of TLER is the problem. I've heard of the TLER toggle tool, but never seen a link to it. If you can find that, I believe KS and RE drives are physically identical, and should be toggleable either way.

As far as WD's concerned, this is by-design behavior. You can RMA the drive if you like, but the next one you get could do the same thing even if it's working properly.

I don't know why that is happening. Can you link to a block diagram for your motherboard? My guess is there are two seperate sata controllers, and it's not possible to stripe across the two.

 
I understand that raid10 would be just as vulnerable to the problem. But it occurs to me I've been previously running raid1 or raid0 on mobo based controllers for years now with desktop drives - never had a problem, makes me wonder about the intel ich8r.

Here is a table from the intel matrix storage manager:

table.jpg


quoted from asus site:
"Intel® P965 / ICH8R with Intel® Fast Memory Access Technology"
"Southbridge
- 6 x SATA 3.0 Gb/s ports
- Intel Matrix Storage Technology supports RAID 0, 1, 5 and 10.
JMicron® JMB363 PATA and SATA controller
- 1 x UltraDMA 133/100/66 for up to 2 PATA devices
- 1 x Internal SATA 3.0 Gb/s port
- 1 x External SATA 3.0 Gb/s port (SATA On-the-Go)
- Support SATA RAID 0, 1 and JBOD"

There are 2 controller but they are seperate. The intel controller has 6 ports to itself. The jmicron has 1 pata, 1 sata and 1 esata. It's a seperate controller bios during bootup.

The table clearly shows support on the ich8r for a 5-6 drive raid5. Yet, when I built the system, I had 6 drives connected to the southbridge's 6 sata ports, all 6 drives showed in the intel raid bios ( I had planned on a 5 drive raid5 and the other drive for a c drive). I created the raid 5 array and was prompted 'please select 3 or 4 drives'. That was it - no option to configure, no settings to change, simply limited to 4 drives.

Anyhow, this hole TLER thing has me sort of depressed about the whole thing. I've wasted a lot of time setting everything up and I can't afford to replace all these drives with RE versions. I didn't buy all of these KS drives just for this, I had sort of collected them over the last year and half through upgrades, portable hd's etc. So when I saw this new mobo with raid5 support I thought I could consolidate - but it seems like wishfull thinking.

I would be happy if I could set it up as JBOD and schedule par2 to run on it everynight and write parity to my mirrored c drive. But I don't think the ICH8R supports JBOD and maybe this is just as vulnerable? What about just using the disk management snapin to do JBOD? Is that a performance hit?

Thanks for your help!

Kandor
 
The problem with raid 5 is the controller must know that the disks are reliable - they're storing parity properly - so that in case of a disk loss, there remains a whole copy of the data. This means it must be paranoid about kicking disks. If it even suspects that a disk might not be working properly (e.g it stops responding for more than 7 seconds) it kicks it. With JBOD, it doesn't need to be paranoid - if a disk dies, the data goes, so there's no need to ensure the reliability of any given disk. So if one disk goes out for 7+ seconds, it can ignore it and hope it comes back, and that's all. Kicking a drive from a raid 5 array is a no-data-loss scenario (at least for the first disk...) but for a JBOD it's immediately a problem. Thus the controller is (or at least, can be!) more lenient with JBOD+KS than raid5+KS.

Hope that explains it. The same idea, by the way, would apply to raid 0+1 or 1+0.

I'll take a look at the manual when it finishes downloading (curse you, asus! your download speeds aren't even funny!) and see if I can figure out why it's not allowing 5/6 drive arrays.
 
I appreciate your help.

Two things I noticed - I download the wd drive fitness dos version and made a bootable cd. I turned off raid temporarily so I could test the drive with this program. It came up and said 'bad cable'. So I thought, heh I have 6 sata drives in here odds are increased I'll have a bad cable (they are all new that came with the p5b mobo). So I took everything apart and looked at the cables - funny thing is there is 28awg wire in these asus cables.

So I looked at my msi cables from before and they are much thicker and beefier - 26awg wire. So I've replaced all 6 cables with 26awg, reran the fitness test (just the quick one for now) and it passed perfectly.

Also, I've checked that all drives in the array have the same firmware.

Kandor
 
Apparently Asus is no longer including block diagrams with their manuals (and therefore I'm no longer buying their boards...) but they mention that ports 3 and 4 are colored red and only to be used as data disks, not boot disks. See page 54 of the manual. Also, on page 147, it specifically says only 2 to 4 disks may be used for raid 5. I don't know how they plan to make a 2 disk raid 5 array (well, conceptually it works, and ends up looking an awful lot like a raid 1 array, but nobody implements that that I know of). Last straw for the camel is Intel's ICH8 datasheet. Page 162 mentions that there are in fact two seperate SATA controllers inside the chip, and page 164 mentions 3 disk and 4 disk arrays, but that's all.
 
But, I read the different part of the manual, obviously :) Thanks for the research. So after all - it seems like the smart thing to do with these drives is set them up as a 4 drive raid 10 were the controller will theoretically be more forgiving?

Kandor
 
Maybe this is not due to the drive-specific issue you mentioned.

I got identical behavior with my RAID setup. I used the Intel RAID to control 4 Seagate 7200.9 80 GB drives. But this was in Vista x86 RTM.

It worked fine for two days then died last night while I was using the system (opening a previously mapped share drive on another computer). It died with a BSD: IRQL_NOT LESS ...

Here are my specs:
- P5B Deluxe
- Core 2 Duo 6600
- 2 GB OCZ Gold memory
- 4 Seagate Barracuda 7200.9 80 GB drives
- I am not overclocking.
- Vista x86 RTM
 
More Info:

When I rebooted to the RAID control screen, it said the drives were all non-raid. I'm really stumped about what to do next. I might forget RAID altogether.
 
I just found an Itel page that gives different instructions for Intel Matrix RAID on Vista than the steps I followed. I will try those and report back here.
 
so I'm very curious...


whats the concensus... does ICH8R support >4 drive raid 5 arrays?


is a 6 drive raid5 array possible?


ive heard mixed.
 
Unfortunately in my experience ICH8R raid arrays are indeed limited to 4 drives.
not only raid 5, but any supported type of raid array are limited to 4 drives.

I had just purchased 8 WD500YS drives hoping to build a nice raid 5 array.on a gigabyte GA-965P-DQ6 (which blatantly advertise a massive 8-sata ports & Raid5 support)
turns out of the 8 ports, only 6 are in fact on the ICH8R - which I could have learned if I took the time to read the ICH8R spec sheet before - and only 4 at once are available to build any raid array)

I must say I'm puzzled, what's the point of having 6 ports ?
And more frustrating this information is nowhere to be found.
 
Hello Kandor -

Did you ever get a resolution to your WD drives coming up "missing" from the Raid5 array? I have the exact same issue on a new PC w/ the Asus P5B-E (ICH8R chipset) and three WD2500KS drives. I found a utility that can apparently turn the TLER on/off, but I'm wary to use this, and don't know if it changes the KS drives anyway. Maybe I should swap the HD's for the YS series?

Any update you can give me would be greatly appreciated.

Jason
 
Hi,

I resolved my issue by throwing out the new thinner 28ga sata cables that came with my asus p5b deluxe motherboard and replacing them with the older style 26ga cables that I had left over from my old system. Since then I haven't had a single problem in the last 4 months or so.

Unfortunately, I have no way to prove that this was the problem. Anecdotal evidence says it is though....

Kandor
 
Hmmm... Sounds worth a shot, I have three cables of a different variety that I can try out - they seem heavier as they are less flexible, I'll check to see if there is any label to tell the gauge of the wire.

Thanks for the reply - Jason
 
You must log in or register to reply here.
Back
Top