Another one with LSI 9240i - starting to fail after a year ..

[DarkLordSilver]

Hi board

I'm just another guy with an LSI 9240i Controller. Everything was running fine during and after the setup of the NAS. This week i wanted to evacuate my data off an old Hitachi SMS100 array with failing PSU, so i went and put all the data on my FreeNAS Array. It's a small supermicro server with Intel Board and two RAIDz Pools.

4x1TB WD Green
4x2TB WD Green

Controller is said LSI 9240i-8 and everything is running in JBOD, as it was suggested by the manual.

I set up the freeNAS quite some time ago, used is rarely and never had any big issues. But since i startet to put large amounts of data to the array coming from my SMS100, there seem to be problems and issues.

So i got very ugly hangs while transferring and went to look at the console.
- Saw a lot of mfi sense errors on one specific disk
- startet to troubleshoot and learned, that:

-- my driver is wrong (shouldn't use mfi, but i wasn't given a choice when installing)
-- my controller is wrong (shouldn't use LSI9240i, but was recommend when i built the system)
-- my firmware is wrong (should use the 9218 IT Firmware)

What i already did:
- tried to evacuate from pool1 to pool2, because the error only showed up in pool1.
-> Result: now pool2 also shows the error, which seems to come above a certain filling level.

- checked said disks with the sense errors externally for SMART errors, since smartctl won't run on the BSD Box.
-> no errors on the Disk

- Updated the FreeNAS from 9.3 to 9.10 stable
-> no changes in behaviour, errors do still come

So here I am and these are my options:

Option 1: I have another RAID Controller, Adaptec 3805. Put this one in and hope FreeNAS will handle the change

Option2: Leave it as it is and scrub both pools and hope it will fix it

Option 3: Flash the LSI controller to 9218 IT and hope it will be ok then.

Option 4 Bonus: What i'm going to do anyway: Try to copy everything off the NAS to a HD before i start doing something.


What would you recommend?

 

[Ericloewe]

-- my controller is wrong (shouldn't use LSI9240i, but was recommend when i built the system)

It's not, it just needs to be crossflashed.

(should use the 9218 IT Firmware)

There is no such thing. You need the LSI SAS 9211-8i IT P20 firmware.

Option 1: I have another RAID Controller, Adaptec 3805. Put this one in and hope FreeNAS will handle the change say goodbye to my data

FTFY

Option2: Leave it as it is and scrub both pools and hope it will fix it

You have serious issues that need to be explicitly addressed.

Option 3: Flash the LSI controller to 9218 IT and hope it will be ok then.

See above for what you should be looking for.

You may have to move the data off the server first, depending on what exactly the controller did to the drives. They might not be readable by a non-mfi controller.

 

[maglin]

It's not the controllers fault. You set up a RAID array behind the ZFS array of what I assume is a single drive vdev and you are reaping the rewards now.

I've read a lot of stuff on FreeNAS and have never once seen anyone suggest to make a RAID array on a RAID controller.

I hope you can offload your data and get your system up to par. Also you need at least 8GB of ram as well.


Sent from my iPhone using Tapatalk

 

[SweetAndLow]

Hi board

I'm just another guy with an LSI 9240i Controller. Everything was running fine during and after the setup of the NAS. This week i wanted to evacuate my data off an old Hitachi SMS100 array with failing PSU, so i went and put all the data on my FreeNAS Array. It's a small supermicro server with Intel Board and two RAIDz Pools.

4x1TB WD Green
4x2TB WD Green

Controller is said LSI 9240i-8 and everything is running in JBOD, as it was suggested by the manual.

I set up the freeNAS quite some time ago, used is rarely and never had any big issues. But since i startet to put large amounts of data to the array coming from my SMS100, there seem to be problems and issues.

So i got very ugly hangs while transferring and went to look at the console.
- Saw a lot of mfi sense errors on one specific disk
- startet to troubleshoot and learned, that:

-- my driver is wrong (shouldn't use mfi, but i wasn't given a choice when installing)
-- my controller is wrong (shouldn't use LSI9240i, but was recommend when i built the system)
-- my firmware is wrong (should use the 9218 IT Firmware)

What i already did:
- tried to evacuate from pool1 to pool2, because the error only showed up in pool1.
-> Result: now pool2 also shows the error, which seems to come above a certain filling level.

- checked said disks with the sense errors externally for SMART errors, since smartctl won't run on the BSD Box.
-> no errors on the Disk

- Updated the FreeNAS from 9.3 to 9.10 stable
-> no changes in behaviour, errors do still come

So here I am and these are my options:

Option 1: I have another RAID Controller, Adaptec 3805. Put this one in and hope FreeNAS will handle the change

Option2: Leave it as it is and scrub both pools and hope it will fix it

Option 3: Flash the LSI controller to 9218 IT and hope it will be ok then.

Option 4 Bonus: What i'm going to do anyway: Try to copy everything off the NAS to a HD before i start doing something.


What would you recommend?

You are trying to blame the manual and hardware guides but non of them say to use a raid controller in jbod. They all say to flash the card to IT mode. A huge hint should be the bad performance and missing smart tests.

 

[DarkLordSilver]

@Ericloewe Sorry, that was a typo. Meant to write the LSI SAS 9211-8i IT P20 FW.
So, but a crossflash will work? I'll evacuate the data for sure (at least what can be evacuated).

@maglin
No I didn't setup the RAID, I passed through as JBOD, as it was recommended in various places. Didn't safe those links for later prove though.
I have 8 devs, not 1 or 2 and they are not raided.

@SweetAndLow
It's been a while since I set up that box, so i can't remember exactly, what recommendation I followed. I guess it's this one here: http://www.freenas.org/blog/a-compl...are-design-part-i-purpose-and-best-practices/

If the RAID controller has a passthrough or JBOD mode, use it

Comming from the official site made it look legit for me.

Perf wasn't that bad after all and the SMART never ever showed signs of not working. That was pissing me off a bit, smart doesn't work but it says everything is fine in the GUI.



Thanks so far for all the replys!

 

[jgreco]

-- my driver is wrong (shouldn't use mfi, but i wasn't given a choice when installing)
-- my controller is wrong (shouldn't use LSI9240i, but was recommend when i built the system)

No, this has never been recommended. We've always said that you can crossflash a 9240-8i to be a 9211-8i.

 

[DarkLordSilver]

Yeah, I know it by now, thanks.

So, i went back to my FreeNAS Server and it won't boot up anymore and hangs while importing the zpools.
It gets a lot of hard errors and sense errors while importing and finally fails with "kdb enter panic" and I'm left with the kdb command line. Well, well, well.

The one thing i keep seeing, is that one disk from the second pool keeps the light flashing, for whatever reason.
My last try for today now hangs with:

spa=ZFS_Pool2 async request task=4

 

[danb35]

(shouldn't use mfi, but i wasn't given a choice when installing)

No, you weren't. The system automatically uses the driver that matches your hardware.

shouldn't use LSI9240i, but was recommend when i built the system

No, it wasn't, at least not by any reputable source.

I guess it's this one here: http://www.freenas.org/blog/a-compl...are-design-part-i-purpose-and-best-practices/

Ah, that would be the one that says:

Therefore, connecting your drives to an HBA or directly to the ports on the motherboard is preferable to using a RAID controller

...and prefaces the quote you selectively gave with "if you must use a RAID controller."

 

[DarkLordSilver]

I know, I wasn't but why would a system even give the option to proceed and why the f was that driver not removed, when it is know, that it caused so much troubles.

Yeah, i had to use, since i owned it already. However, I think this has been discussed more than enough now, right? No wonder this forum is slandered in the interwebs as full of hardware nazis.

But to be helpful to the next victim of FreeNAS. The resolution is the following;
If the Volume doesn't come up again, observe on which disk the error is thrown, then shut down the server after the kernel panic arrives. Plug out the disk. Restart and evacuate the data.
Disk scrub should work as well again, but on the other hand, I'm just trying that on the first volume for fun and to see if the zfs can really recover.

 

[jgreco]

I know, I wasn't but why would a system even give the option to proceed and why the f was that driver not removed, when it is know, that it caused so much troubles.

Because for certain applications, it works fine.

Yeah, i had to use, since i owned it already. However, I think this has been discussed more than enough now, right? No wonder this forum is slandered in the interwebs as full of hardware nazis.

Yes. We're hardware nazis. There's no effing reason for it. We're just totally arbitrary about it and we're just trying to piss you off. It's not like there's any actual risk involved in using the hardware that we so strenuously advise against you using. Not like you'll ever end up here as a result of problems experienced.

It couldn't possibly be that we actually want people to have successful experiences and that we've seen too many disasters where people can't be arsed to follow the recommendations and then had it all go sideways, so we do everything we can to make sure that people follow the recommendations and work with successful hardware recipes.

Thread closed.