X9SCL-F with E3-1220v2 and 4x4GB memory hangs under one specific load

blacklion

New Member
Jan 10, 2017
5
0
1
41
I have home storage server, which was upgraded recently to X9SCL-F MoBo, E3-1220v2 CPU and 4x4G RAM modules (Kingston KTH-PL316ES/4G). MoBo is flashed with latest IPMI firmware (3.38) and BIOS (2.20).

PSU is Seasonic G-450 (I know, that it is not server PSU, but it is home server with «desktop»-style case!). Storage subsystem is 5x3TB Toshiba DT01ACA300 HDDs (for data, at SATA ports 2-6) and Samsung 750EVO 240G SSD for system & ZFS L2ARC on SATA port 1. No additional controllers are installed.

All this hardware (case, PSU, storage) is transferred from previous version of this server, which was build arounf P45-based MoBo and Q9550 CPU. It worked flawlessly with this «old» hardware

This «new» system works well until I run CrashPlan backup client (which I use for many years already).

When CrashPlan is running, system hangs completely after some time (from several tens of minutes to several days). I mean, it is not OS crash or hang-up, when kernel debugger works. It is complete hardware hang-up. Even «Reset» button or/and IPMI command doesn't work (first in my life!). Only power cycle is able to reset system. IPMI works, though.

When system is hanged up, IPMI shows that all voltages are Ok, temperature is Ok, everything is Ok. Seconds before hang up CPU temperature reported by OS is like 32-35°C, which looks great. HDDs are perfectly healthy and SMART doesn't show any errors.

I could not reproduce this hang-up with any synthetic load: read all data from HDDs and compiling system with a lot of parallel jobs at same time don't trigger this behavior.

Memory looks good, as I've run memtest86 (free version) for 8 passes (10 hours) without any problem.

PSU power should be enough, as it runs flawlessly with "hotter" CPU and MoBo (old P45-based Mobo with external NIC and Video and Q9550 CPU).

Is it here something known about this MoBo what could cause such behavior? Maybe, I've misconfigured BIOS in some way? Maybe, here is something what can I try to diagnose this problem?

Unfortunately, I don't have resources to try other Mobo with this CPU and memory.

I'm using FreeBSD, but it doesn't look like software problem.

I've read thread about this MoBo has problems with additional ROMs, but I don't have any problems with booting and I don't have additional controllers. Also, BIOS 2.20 is newer than one which has problems.
 

cheezehead

Active Member
Sep 23, 2012
711
173
43
WI
Try running memtest on it and see if anything pops there. The config should work assuming there's no hardware issues I would think.
 

dirk adamsky

New Member
Apr 24, 2012
14
0
1
Hi Blacklion,

I have had similar experiences with x9-scm-f boards with Proxmox (linux based).
Turned out that one of the nics caused the OS to hang completely (driver problem?).
My collegue and I were able to reproduce the hanging by sending a large file (> 10GB) from that machine.
Our solution was to add an extra dual port Intel nic to the board and only use that nic.
After that the board was stable as a rock.
We have also used the x9-scm-f board with omnios+napp-it without any problems with the onboard nics (but that is Solaris/Illumos based).
Unfortunately i do not know if the x9-scl-f has the same onboard nics.
 

PigLover

Moderator
Jan 26, 2011
2,954
1,262
113
The X9SCL-f has two different models of Intel NIC. One of them (The 'M' designation) is known to have a number of bugs that can lead to symptoms like you are having. Since the workload that triggers this is both Network related and repeatable it might be worth a shot at swapping the ethernet to the other port to see if it clears up the problem.

BTW, this is why SM released an updated version of this MB - X9SCL-+f - with both NICs using the "L" version of the chipset (also because VMWare initially refused to release drivers for the buggy NIC). When these boards were current SM would swap for the newer board under warranty. But since we're probably 5 years or so past their introduction they probably won't do it now - but then again, can't hurt to ask them!
 

blacklion

New Member
Jan 10, 2017
5
0
1
41
Try running memtest on it and see if anything pops there. The config should work assuming there's no hardware issues I would think.
I've run it for 10 hours. No problems.

The X9SCL-f has two different models of Intel NIC. One of them (The 'M' designation) is known to have a number of bugs that can lead to symptoms like you are having.
It is interesting. Really, here are 82579LM and 82574L and, yes, I'm using 82579LM now. I'll try to switch to 82574L as I don't need tow NICs. But I don't hold my breath, because this system allow me to read and save gigabytes of files via network (it is my NAS!) when CrashPlan is not running. But maybe, something in CrashPlan protocol triggers subtile bugs in this chip. Thank you!
Board was bought not from SM distributor so, I think, replacement is out of question :)
 

cheezehead

Active Member
Sep 23, 2012
711
173
43
WI
FWIW, i'm running an X9SCL for my pfSense box (using both onboard nics) and have had no issues.
 

Rand__

Well-Known Member
Mar 6, 2014
4,423
863
113
Can't hurt to mail SM support re this, usually they are happy to help.
 

blacklion

New Member
Jan 10, 2017
5
0
1
41
Nope, switching to 82574L didn't help. I'll contact SM support, but looks like it is broken motherboard.
 

blacklion

New Member
Jan 10, 2017
5
0
1
41
For the record:
  1. it was PSU.
  2. SM support said, that product EOL, and they could not provide any support for it.