Fun Times with Installing 9.3 on a SuperMicro X10SLL-F

[DrKK]

Edit: Cross-linking with this potentially similar/related issue with the same board.

So, I sort of know what I'm doing. And I just had some interesting times building out a FreeNAS 9.3 with the X10SLL motherboard. As you guys know, the SuperMicro X10 series is what we recommend for a 1150-based home build. It is the go-to board. The difference(s) between the X10SLL and X10SLM series are very minor---the chipsets are the same, there is a small difference in the PCI slots (which I don't need for anything), and there is a small difference in which USB ports are 2.0 and 3.0 and how many there are. Other than that, the boards are pretty much identical. The X10SLL has a USB 3.0 type A female mounted right on the motherboard, which is kind of interesting.

Anyway, the problems began straight out of the gate.

I have the G3250 CPU bought for this motherboard. I figured; ok this has been out long enough, the BIOS on a newly shipped board should be ready for this.

Well it wasn't. The v1.1 BIOS could NOT handle any of the recent Haswells, and expected the old school stuff, which is (alas) getting pretty damn hard to find. Micro Center, for example, doesn't even model for it in their inventory now. So, update the BIOS right? Again, no dice. While the v2 BIOS does allow more 1150 socketed CPUs, it turns out, the G3250 (and many other newer CPUs in the 1150 socket) do not work. This has been confirmed with the SuperMicro engineers.

So yeah. Wasn't happy about that. But I did find a G3220 on Amazon, so two days later I had that in my hands. Now, the board POSTed, and everything was fine on that score.

I then plugged in an 8GB Kingston Micro DT (again, the default, go-to USB boot device that we recommend for FreeNAS) into one of the backside USB 2.0 ports. It was recognized immediately as da0, and the FreeNAS install CD put 9.3 right on there. Again, no problem (except for a scary moment when it appeared to just be sitting there for awhile--you have to be patient).

But, then I mountroot'ed, super hard, on the reboot into the installed OS. Could not get past the mountroot. As far as I could tell, everything on the boot drive was correct, correctly partitioned, and ready to go, so I really had no friggin idea why the recommended hardware, from soup to nuts, was giving me a mountroot.

So I started crying to Cyberjock, who began troubleshooting with me on the Mumble server, but then he had to leave and go to the store. So I began to get more upset.

I figured, OK, it can't be any worse than this. So let's start screwing around. I reset the BIOS to the defaults. No change. I cold boot and clear the CMOS and all that jazz. No change.

But then I pull the USB drive out, and put it into the USB 3.0 BUILT-IN MOTHERBOARD usb port (i.e., in the middle of the motherboard, not on the back panel), and I was quite certain that ought not to work for crap (since nothing is supposed to work, really, in USB 3.0 ports).

Well I'll be damned if it didn't boot right up, and everything was fine.

I took it out, put it back into a USB 2.0 port on the back panel. Mountroot. Tried another port. Mountroot, but it seemed to get a little farther.

Put it back onto the USB 3.0 port inside the motherboard? Boots fine.

So I can't explain it.

If anyone out there is using recommended hardware, especially a SuperMicro X10 series, and is getting a mountroot with the 9.3, try putting your boot device in another port. Keep trying all the ports.

Everything is working fine with that new build now. It's quite mysterious. I intend to deliver this build to the guy that asked me for it, and perhaps we'll buy another X10SLL and play around a bit and try to understand exactly what the hell is going on, and make a follow up post. Maybe.

I hope this was useful information to someone going crazy out there wondering why they are having mountroot problems in the X10.

 

[solarisguy]

Thank you!

That was very informative, and is actually changing my immediate plans.

 

[Ericloewe]

This is quite interesting.

I'll note that there are two "families" of X10 server boards, from what I know:

X10SLL-F and X10SLM-F

X10SLL+-F, X10SLM+-F, X10SLM+-LN4F and X10SLH-F (and possibly the X10SL7-F, but that one's a bit different)

On my X10SLM+-F, I've always had my boot device (still 9.2.1.9, haven't had the time to update) on the internal USB 3.0 type A port and it's always worked fine. Next time my server is on the desk bench for maintenance and dual boot devices, I'll be sure to investigate the other ports and see what happens.

As for the G3250 not working, ark.intel.com says it was released in Q3 2014, unlike most Haswell Refresh stuff, which was released in Q2. My guess is the new model numbers weren't added to the BIOS in time (Haswell and Haswell Refresh are the exact same silicon) and they haven't bothered (yet?) to issue another update.

 

[DrKK]

So, the only thing I can think of:

I assume the USB ports on the board (just because there's like half a dozen, plus the headers for any panelmounts) are somehow controlled by different controllers. Something about the timing, the primacy between ports for boot sequence, or the connection to the southbridge, or something, is causing something that's supposed to be ready-to-go at instant N to be delayed by epsilon. Inverting some kind of temporal relationship some kind of device behavior has to have or something. This is not really my strong suit, so it's probably all horseshit. But it's the only thing I can think of.

None of this seemed to be a problem with the old (pre 9.3) boot-loader, which was a wholly less complicated scheme. So my hypothesis about all the pieces falling into the the right place at the right time, or it won't boot, seems to be possible.

What we need is someone with enough free time on their hands with this problem to find out if the problems go away when 9.2.x is installed vice 9.3. Then we will be quite certain it's the 9.3 bootloader rubric, and the nerds over there at iX might be deployed accordingly.

 

[Ericloewe]

All USB ports should be handled by the PCH, so I'm not too sure about timings. I'd bet on a BIOS bug of some sort, or a hidden "feature" that expects boot disks to reside on internal/header USB ports, instead of back panel ports.

I won't have time to investigate until February, but I'll make sure to do so thoroughly ASAP.

 

[DrKK]

Do we have a relationship with the SuperMicro guys? I bet some large percentage of their sales are to people building FreeNAS's. It'd be nice if they could ship us hardware for diagnosis of anomalous behavior, and/or provide us with software and/or certain not-so-easily-found specs (like, complete lists by SKU of what boards support what cpu's under what BIOS revisions) for the community.

After all, we're playing for the same team.

 

[mjws00]

Heh. Large percentage of their sales or "The bloody clueless noobs that keep buying server gear and have no business inside a case." ;) ;)

Interesting experience DrKK, thanks for writing it. Did an SL7 a few days ago from newegg. Came with latest BIOS and BCM. Unfortunately no interesting new CPU to try in it. I'll throw the USB's in all the ports just for kicks. I did boot internally, externally, and virtual via ipmi nothing of note happened.

 

[DrKK]

Heh. Large percentage of their sales or "The bloody clueless noobs that keep buying server gear and have no business inside a case." ;) ;)

Interesting experience DrKK, thanks for writing it. Did an SL7 a few days ago from newegg. Came with latest BIOS and BCM. Unfortunately no interesting new CPU to try in it. I'll throw the USB's in all the ports just for kicks. I did boot internally, externally, and virtual via ipmi nothing of note happened.

you know what would be good, is a sticky post, delineating the differences among the X10 series.

I mean, with some work and research, people can see the differences in NICs and controllers and ports and blah blah blah, but to have a nice convenient sticky or something, so that everyone at once knows the salient difference between an X10SLM, SLL, and SL7, and what not, would be good juju.

 

[Ericloewe]

you know what would be good, is a sticky post, delineating the differences among the X10 series.

I mean, with some work and research, people can see the differences in NICs and controllers and ports and blah blah blah, but to have a nice convenient sticky or something, so that everyone at once knows the salient difference between an X10SLM, SLL, and SL7, and what not, would be good juju.

"So, you've decided to steal cable purchase a Supermicro X10 motherboard"

 

[DrKK]

"So, you've decided to steal cable purchase a Supermicro X10 motherboard"

I don't get it? :(

 

[mjws00]

I don't get it? :(

Heh. Took me a second as well. Very witty, sir.

 

[Ericloewe]

I don't get it? :(

I'll let the video do the explaining.

 

[Ericloewe]

you know what would be good, is a sticky post, delineating the differences among the X10 series.

I mean, with some work and research, people can see the differences in NICs and controllers and ports and blah blah blah, but to have a nice convenient sticky or something, so that everyone at once knows the salient difference between an X10SLM, SLL, and SL7, and what not, would be good juju.

https://forums.freenas.org/index.ph...d-to-buy-a-supermicro-x10-motherboard….25951/

 

[mjws00]

Just for kicks. I did try all the ports on my SL7. All worked. All external. The internal female. And ports off the header. Using 9.3 stable. Bone stock BIOS.

 

[danzi]

Oh oh my... I just posted a build plan with this board. Do I get to rewrite that plan?? :(:(

 

[danzi]

Had some e-mails back and forth with Supermicro's tech team regarding the CPU/RAM with the X10SLL-F... Just thought I'd share it here:


---------------------------

The X10SLL-F with BIOS version 2.0+ can support all socket 1150 CPU’s in the Celeron, Pentium i3 and Xeon E3 families, the i5 and i7 are both not supported.

Also, you must use ECC enabled memory, although this will only work on the E3, the system will not POST if non-ECC memory is used. This is due to a chipset limitation.

I might miss on specific features of cpecific CPU’s, if you would have an ECC enabled i3 CPU then ECC should simply function as expected.

As for memory, we always recommend to use something from the tested memory list, but we haven’t heard many complaints about memory not working except for when it is not ECC memory, the board isn’t too picky about it.

Met vriendelijke groet / Best Regards / Mit freundlichem gruß,


Quintijn Schade van Westrum

Application Engineer

Super Micro Computer B.V.

Technical Support

 

[Ericloewe]

Interesting, so it might be a bad processor, if they say they're all supposed to work.