Seagate 3tb (7200.14) doesn't work with hardware raid?

[titan_rw]

Posting this to offtopic since it's not really freenas related.

I had 6 seagate 7200.12's (1tb) drives in hardware raid 5 on a highpoint rocketraid 3520 card. This worked absolutely awesome from 2008 or so until recently. A bunch of the old drives started to develop bad sectors. I didn't want to invest in new 1tb drives, so I thought I'd upgrade the whole array. I had some existing 3tb barracudas (the 7200.14 desktop ones), and bought a couple extra to make 6.

I wanted some fast local storage for the machine (windows workstation), so I arranged the 6 as raid 10 (3 sets of 2 way mirrors). (this array has always had selective backups run daily, which are sent to my secondary freenas machine). Twice in the first week I had the raid card 'fail' drives out of the array. I'm 95% sure the drives are fine, so I'm thinking it's something similar to cyberjock's experience with seagate 'desktop' drives.

What's funny, is that these exact same drives (7200.14 desktop) work fantastic in freenas and ZFS. I've got an 11 disk z3 in my main nas that's been completely problem free. I guess it's the hardware raid controller being more 'picky'? Zfs is more forgiving about the quirkyness of desktop drives?

So I think I'm just going to turn these 6 seagates into a zfs z2 pool. I've got some WD Blacks that should work fine on the rocketraid instead.

Anybody have other experiences with 7200.14's on a hardware raid card?

 

[pirateghost]

This sounds like a question for the rocketraid folks. Can that card handle >2tb disks?

 

[cyberjock]

The 3520 should be able to handle >2TB disks (I had the 3560) as long as you have the latest firmware.

I will warn you that I bought $2000 worth of Seagate disks in 2009, then they got pulled about 120 days later. There was a deliberate change to the firmware starting with that generation to make the disks far less friendly for RAID configurations. In my case I was randomly dropping drives so fast that I couldn't rebuild the array before I'd lose a second one. I had hardware RAID6 and I had quite a few nights where I was sweating that I'd lose the pool. Nothing was wrong with the disks and Seagate's official answer was "Those are desktop drives and aren't meant for RAID arrays.". There was a 30+ page thread in their forums on the issue and they'd say nothing else except "Use the disks as they are intended.. k, thanks". The forum thread pretty much speculated that Seagate had determined they were losing too much money due to the desktop drives cannibalizing the enterprise models and they wanted to make sure this wasn't happening.

It had been about the 90 day mark before I started having problems, so return to newegg wasn't an option. I couldn't RMA them because every diagnostic you'd run on them said everything was fine. Ended up switching to all WD drives and suddenly the problem went away. All those disks still sit in a box to this day as I don't trust them at all.

After that, I say screw Seagate. I'll never go back to them after spending $2000 for the "opportunity" to use their drives for 3 whole months. The only advice I can give is to either bail on the hard drives or bail on the controller. The disks have problems that cause most controllers to drop them, it's not just our family of cards or our brand(trust me.. I've tried everything and they just suck for anything except desktop use). They do work on standard SATA controller from my very limited testing. So if you want to use them in a desktop you should be okay. Note that I say should because I'm not a fan of Reinstalling Windows every week because a hard drive deliberately sends HALT commands to the SATA controller in an attempt to get RAID controllers to drop the drives because they appear to be failing.

I'm sure you're thrilled to hear my answer. Be glad you didn't drop $2000 on 14 disks only to find them useful as paperweights 4 months later. Before that whole fiasco with Seagate, I swore by them. Every machine I bought had Seagates. Every friend or family that asked what to buy I recommended Seagate. I'll never recommend Seagate under any circumstances again for the rest of my life. I can't be dropping 4-digit dollar amounts on disks that can't function in whatever capacity I want to use them in. Until Seagate wants to give me $2000 worth of disks I'll never even consider trying them again(yea.. like that'll ever happen). I used to be responsible for 100-200 disks a year, so I'm glad I left them as WD has definitely enjoyed my business. Their RMA website is easier to use, their RMA process itself is faster, etc.

 

[titan_rw]

Yea, the controller seems to support the 3tb drives ok, as it is running the latest firmware.

I definitely feel for cyberjock's loss of $2000. If I remember right, we figured out they were 7200.11 models? And I had 6 7200.12's in hardware raid 5 for 4-5 years with zero trouble. Even did a few disk replacements. It's weird that the 7200.11's would be that different.

I wonder if my problem with the 7200.14's if it is something specific to the rocketraid controller. I do have a m1015 spare that I could flash to 9211-IR mode, and try raid10 on it. Maybe it would handle the drives better. I thought the rocketraid would have been the better controller as it has on board cache (only 256 meg though), and is a 'higher end' controller in some ways. Although it is an older higher end.

Like I said, with ZFS, I've had no problems at all with the 7200.14's Asside from a slightly higher infant mortality, they seem quite reliable.

 

[titan_rw]

Definitely seems to be a conflict with the rocketraid and the seagate drives. If I hang the same 6 drives off of a m1015 in IR mode, it seems to work just fine. Now it seems I need to flash it to 9240 mode to be able to change stripe size, which seems to be locked to 64k in 9211 mode.

 

[cyberjock]

Definitely seems to be a conflict with the rocketraid and the seagate drives. If I hang the same 6 drives off of a m1015 in IR mode, it seems to work just fine. Now it seems I need to flash it to 9240 mode to be able to change stripe size, which seems to be locked to 64k in 9211 mode.

They'll still drop the disks unless the M1015 has a timeout that is extremely long. I'm not sure what it is set to or if you can change it as I flashed mine to IT mode. But I tried 4 different families of Highpoint,2 families of 3ware, and 2 families of Areca and the drives dropped regularly on all of them.

 

[titan_rw]

I'm starting to think the highpoint card is having problems. Specifically one port on it.

The seagate 7200.14's are running on the m1015 with no problems at all.

Back on the highpoint, I hooked up 4 WD Blacks in raid10. The next day, I had the same port fail on one of the WD's. I was using ports 3-8 for the seagates, and ports 5-8 for the WD's. Port 6 in both cases seemed to be a problem. I did a 'live' port change with the degraded WD array. I pulled port 6, and used port 3, that was previously not used. Started an array rebuild, which seemed to complete ok. So far no problems now that I've avoided port 6. I'm pretty sure it's not the cables, as I've had at least two different sff-8087 breakout cables hooked up.

I don't really need all 8 ports on this card right now, so I'm content to avoid port 6. The card was bought in 2008/9 anyway, so I'm sure it's long out of warranty.

 

[cyberjock]

Are you able to access the SMART parameters and run SMART tests on your disks? That would be a go-nogo test for me as to whether I'd use the card long term if it is just a port issue. I couldn't get SMART to work with Highpoint. I had to go and use the highpoint cli(not included with FreeNAS and doesn't work out of the box). See https://bugs.freenas.org/issues/1932 from 10 months ago when I tried to get it working and requested support be added.

But considering that the current issue appears to be that highpoint support doesn't work with some cards with smartctl the issue is that highpoint has not released their designs to allow smartctl to be written to work with all of their cards(any?). I did give access remotely to someone back in April so they could troubleshoot the highpoint smartctl support with a test system with no actual server data. He logged into the server for 2 hours, then logged out, and has not responded to any emails since. Kind of wondering what happened to the guy.

 

[titan_rw]

All I've ever had work SMART wise on the highpoint was their web interface. And it only gives you access to the 'normalized' values for smart. No raw values, and as far as I'm aware, no capability to run self tests on the drives. I do have email alerts setup in the web interface, and do get emails when the controller detects (and corrects) any bad sectors it finds on the drives. The web interface does give a raw value for 'reallocated sectors', which I assume it pulls from the smart attribute.

I do set regular array verification tasks in the web interface. As far as I know, this does a read verify surface scan.

I'm not too worried as I have the NTFS array backed up to freenas daily anyway.

On another note, I tried flashing the 9240 firmware back to the m1015. The flash worked fine. Was able to enter the new web bios thing, and setup a new array config. Downloaded the lsi drivers for the 9240, but they wouldn't load in windows. All I got was 'code 10, the driver cannot start'. Couldn't figure out how to make it work. Flashed it back to 9211, and it worked just fine. No driver issues. It does seem that 9240 mode gives more options when your intent is hardware raid. I'm contemplating getting a m5014/5 and trying out 9260 firmware. I'm not sure why I couldn't get the 9240 drivers to work.

 

[cyberjock]

All I've ever had work SMART wise on the highpoint was their web interface. And it only gives you access to the 'normalized' values for smart. No raw values, and as far as I'm aware, no capability to run self tests on the drives. I do have email alerts setup in the web interface, and do get emails when the controller detects (and corrects) any bad sectors it finds on the drives. The web interface does give a raw value for 'reallocated sectors', which I assume it pulls from the smart attribute.

Don't trust the webGUI and/or emails to save you. I did that on my Areca controller and I had several disks with problems and I couldn't have proven it if I hadn't tried a different controller and happened to notice the smartctl warnings. I got a very long post somewhere where I broke it all down and the end result was "don't trust anything except the raw values and test results from smartctl". I could have had serious problems if I hadn't caught it when I did because I had more disks with problems than I had number of redundant disks. Whoops!

I do set regular array verification tasks in the web interface. As far as I know, this does a read verify surface scan.

If its a hardware RAID then it just verifies data against parity and fixes whatever it thinks is broken.

On another note, I tried flashing the 9240 firmware back to the m1015. The flash worked fine. Was able to enter the new web bios thing, and setup a new array config. Downloaded the lsi drivers for the 9240, but they wouldn't load in windows. All I got was 'code 10, the driver cannot start'. Couldn't figure out how to make it work. Flashed it back to 9211, and it worked just fine. No driver issues. It does seem that 9240 mode gives more options when your intent is hardware raid. I'm contemplating getting a m5014/5 and trying out 9260 firmware. I'm not sure why I couldn't get the 9240 drivers to work.

No clue how to help there. I reflashed my M1015 and used it in FreeNAS only.

 

[cyberjock]

Found my thread.. it applies to an old Areca controller I had, but the issues with Highpoint are identical. I swore by Highpoints until 2010. I used to recommend them regularly to friends, but now I recommend FreeNAS with ZFS. But I just realized you gave me feedback on that thread, so I'm preaching to the choir.. LOL.

http://forums.freenas.org/threads/areca-raid-controllers-and-smart-support.12694/