Kernel Panic on Supermicro SSG-6048R-E1CR36L

[lordtangent]

I'm getting Kernel Panics when anyone writes to the NAS. Instant crash.

Is it true that the LSI 3008 controller has problems with FreeNAS? I noticed some reports to that effect but there doesn't seem to be specific info backing it up. What is the actual problem? Is it Kernel panics? Is there any way to enable saving the core dumps from the kernel?

Here is our configuration:

FreeNAS-9.3-STABLE-201512121950

Supermicro SSG-6048R-E1CR36L
SAS3 via LSI 3008 controller; IT mode
Dual 10GBase-T LAN w/ Intel X540

2 x Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz
8 x Micron 16MB DDR4 SDRAM RDIMM MTA36ASF2G72PZ

HGST HDN724040AL hard drives

Any recommendations? If we can just change out the HBA that would be a reasonable quick fix. I'd like to use FreeNAS but this is sucking up a lot of time to debug. A quick fix is what I'm looking for here.

Illumos/OmniOS with Napp-it was working flawlessly. But I wanted an easier way to manage AD integration and SMB3 via Samba rather than the Illumos kernel driver (no SMB3 there), hence the the decision to give FreeNAS a try. Kernel panics in FreeNAS are obviously a show stopper unless there is an easy workaround.

Thanks for any help.

 

[Ericloewe]

Is it true that the LSI 3008 controller has problems with FreeNAS?

The latest version should be at least "reasonably stable", possibly "stable enough for production use", but nobody's actually confirmed that it's the latter. Insta-crashes are not expected at all, at this point.

A copy of the debug file from System ‣ Advanced ‣ Save Debug would probably contain clues for such a repeatable experiment.

 

[lordtangent]

Here is the debug file.

 

[cyberjock]

I've heard people claim that LSI 3008 is "stable", but I keep seeing people saying it doesn't work for them. So I disagree and think that the issue *is* the 3008 controller and/or drivers.

I will say that a FreeNAS that is idle should be cooler than 52C by CPU temp. 52C isn't a dangerously high temp, but for being idle that's quite a bit warmer than I'm used to. I'm used to 30-35C idle.

The box isn't crashing, so it is not panicing.

But, the logs show various weird issues. AD is supposedly enabled, but it isn't working properly based on samba logs saying its unable to fetch records. So there's definite configuration issues at the slightest.

 

[lordtangent]

Here is an update.

The pool we were using was built on OmniOS. (This is a detail I omitted in my initial post.) It didn't occur to me this might be part of the issue until a coworker suggested it might be a contributing factor. So I destroyed the pool and rebuilt it in FreeNAS. Bingo. Everything now seems stable. AD is still complaining a bit in the logs but it does seem to work regardless.

The kernel panic seems to be the result of a really odd edge case bug from the imported pool. (We'll let someone else upstream worry about that. I really don't care as long as my immediate problem is fixed) Now that I'm using the FreeNAS generated pool, I think we can label this resolved. Not getting the kernel panics any more.

As a nice side effect of all the debugging steps I went through is my firmware and bios are all now totally up to date. Hopefully I wont need to monkey around with this particular box for a while.

 

[cyberjock]

I'm very confused then... because your debug shows no crash dumps at all. So something must have been going very, very sideways that made crashdumps not work. Crash dumps are supposed to work in all but the must egregious of situations. In any case, glad you got it figured out.