Reboot loop commences after approx 10 mins of running

[Dainis Berzins]

I have a fresh 9.3 installation downloaded this weekend. After installation, the webGUI appeared and I could log in and configure FreeNAS as I would expect. After about 10 mins, I notice that the server appears to have rebooted and is stuck in a reboot loop.

The reboot loop is so short that I do not even get the boot log and the “Press F2 for BIOS settings”. The boot lasts less than 1 second. The power light flashes once, the HDD flashes once as the fan can be heard starting and then the lights go out and the fan stops. Two seconds later, this happens again and continues endlessly
(or at least all night).

I am able to interrupt this by disconnecting the power to the server for several minutes (5-10). Then, sometimes (not always) the loop does not occur and the machine boots and all works as expected until a reboot loop kicks in after about 10 mins. I believe that leaving it disconnected from power for a longer time increases the likelihood that it boots successfully.

Things I have tried:

• Disabling Hyper Threading support in the BIOS.
• USB boot drive installed internally and externally.
• Disabled all other drives in the boot sequence.

I will check the BIOS version of the motherboard later, although it is newly purchased so it should not be too old.

Any ideas what might be causing this?

 

[enemy85]

What's your hw? It looks like sone hw incompatibility

 

[Dainis Berzins]

HW is listed in the signature.

 

[enemy85]

From tapatalk on mobile phones u don't see any signature...

 

[cyberjock]

Well, either you have:

1. A zpool that is corrupt causing the system to crash when the pool is automatically mounted on bootup
2. Some problem with the boot device
3. Hardware that is failing or otherwise incompatible.

About 99.99% of the time, reboots without actually crashing is because of #3. So you need to find out if the machine is crashing or just rebooting.

Signatures don't show up to people using Tapatalk, so listing it in your sig doesn't help the people that are trying to help you.

I would *never* recommend that board. But, a handful of people are using that board successfully (at least, if there's a problem that is causing reboot loops they haven't complained) so I'd say the hardware is capable of working with FreeNAS.

 

[Dainis Berzins]

Well I didn't know Tapatalk hid the signatures. My hw is as follows:
FreeNAS 9.3 64-bit
Motherboard: Asus P8C WS
CPU: Intel Core i3-3250
RAM: 16 Gb DDR3 EEC RAM (2 x KVR16E11/8I),
Volumes / HDD:
- 1 RAIDZ1 volume with 3 x 2 TB WD Green drives
- 1 mirrored volume with 2 x 250 GB Samsung drives
- 1x 3TB WD Red drive.

I reused the PSU from my previous FreeNAS server, but I did not check the power rating to support the new mobo. Perhaps that could explain the behaviour. I'll post an update when I have checked whether it has sufficient power. I can't imagine why else it would be compatinle for 10 mins then need a 10 min break.

 

[Bidule0hm]

To me it's the current peak from the drives during the spinup who causes the voltages to sag and the MB to reset to protect herself because of that.

Try with a better PSU or less drives powered if you don't have another PSU.

 

[Ericloewe]

What PSU are we talking about?

 

[Dainis Berzins]

Power supply is a Zalman ZM460B-APS, 460W.

I have now tried unplugging the drives, first some, then all of them. The reboot loop still occurs. So I think I can rule out the lack of power from the PSU.

But I have found what I think could be causing this. The Core i3-3250 processor is listed as supporting ECC memory in the product specification, which is why I went with it. However, the datasheet vol 1 (page 23) only refers to non-ECC memory. Both sources are published by Intel. So my guess is that the processor does not actually support my ECC memory. I have contacted Intel for clarification as to whether ECC is or is not supported.

So, cyberjock, I'm betting you were right about hardware compatability being the likely cause. Thanks!

 

[Ericloewe]

The CPU supports ECC according to ark.

That PSU seems to be a crummy design - and it certainly is an ancient design.
The only way to rule it out is to try out a known-good unit, unfortunately.

 

[DrKK]

Also, given the timing, and the transient nature, but inescapable nature, what came to my mind first was a thermal issue. Is there any chance that heatsinks/whatever on the CPU and/or northbridge (if it has one) are not firmly seated?

 

[BigDave]

Have to agree with the DrKK on this one, check the UEFI for hardware voltages and temps first.

 

[Dainis Berzins]

OK, various things for me to look into. Thanks.

But regarding the CPU, ARK does say ECC is supported which is what I checked before buying the CPU, but the datasheet link in the ARK product spec page leads to theDesktop 3rd Gen Intel® Core™ Processor Family: Datasheet, Vol. 1. Page 23 only lists options for non-ECC memory:

• DDR3 / DDR3L at 1.5 V Data Transfer Rates
  • 1333 MT/s (PC3-10600), 1600 MT/s (PC3-12800)
• DDR3 / DDR3L at 1.5 V SO-DIMM Modules
  • Raw Card A – Dual Ranked x16 unbuffered non-ECC
  • Raw Card B – Single Ranked x8 unbuffered non-ECC
  • Raw Card C – Single Ranked x16 unbuffered non-ECC
  • Raw Card F – Dual Ranked x8 (planar) unbuffered non-ECC
• Desktop platform DDR3/DDR3L at 1.5 V UDIMM Modules
  • Raw Card A – Single Ranked x8 unbuffered non-ECC
  • Raw Card B – Dual Ranked x8 unbuffered non-ECC
  • Raw Card C – Single Ranked x16 unbuffered non-ECC

Page 12 also says "Desktop PCH platforms support non-ECC UDIMMs only". So does this CPU really support ECC and if so, what memory configuration is supported? That's a question that I have posted to an Intel community forum as it is not a FreeNAS issue, but if anyone can reconcile these conflicting details, I'd appreciate it.

 

[Ericloewe]

They only support ECC with C22x PCHs, not with Zxx/Qxx/Pxx/Bxx/Hxx chipsets.

 

[Dainis Berzins]

Intel have come back to me in this forum post to indicate that the Core i3-3250 actually does not support ECC memory at all. They admit that their ARK information was incorrect. Seems like I wasted my money because of Intel's wrong information.

So I am now awaiting the delivery of a Xeon processor that they say should support ECC memory.

 

[Ericloewe]

Intel have come back to me in this forum post to indicate that the Core i3-3250 actually does not support ECC memory at all. They admit that their ARK information was incorrect. Seems like I wasted my money because of Intel's wrong information.

So I am now awaiting the delivery of a Xeon processor that they say should support ECC memory.

Ok, that is important information that leaves me somewhat uneasy, since all Ivy Bridge i3s were supposed to support ECC. At least they quickly fixed the ark page.

 

[DrKK]

Wow! Well I'll be damned.

So that's interesting, no? The motherboard would POST fine, and everything would appear to be fine for ten minutes. I'm surprised you even got that far if the board doesn't support ECC.

Learn something new every day!

This has my vote for thread of the week.

 

[Dainis Berzins]

The board does support ECC if a Xeon E3-1200 V2 family processor is used, according to Asus and the small print in the QVL for memory for the board.

I still have a question with Intel regarding their assertion that the C216 chipset (that my mobo uses) only supports this Xeon family of processors. Asus support tell me they have tested and verified many other processors that they list in their homepage (including the i3-3250 that I had bought).

But the answer to this question doesn't solve lack of ECC support for my i3-3250 and thus the need to switch to Xeon.

 

[Dainis Berzins]

Intel have published the list of Celeron and Pentium processors that support ECC memory in their community discussion forum. They have/will correct the ARK information:

• G1850
• G1840T
• G1840
• G1830
• G1820TE
• G1820
• G1820T
• G1630
• G1620T
• G1620
• G1610T
• G1610
• G3220
• G3220T
• G3240
• G3240T
• G3320TE
• G3420T
• G3420
• G3430
• G3440
• G3440T
• G3450

 

[cyberjock]

The board does support ECC if a Xeon E3-1200 V2 family processor is used, according to Asus and the small print in the QVL for memory for the board.

I still have a question with Intel regarding their assertion that the C216 chipset (that my mobo uses) only supports this Xeon family of processors. Asus support tell me they have tested and verified many other processors that they list in their homepage (including the i3-3250 that I had bought).

But the answer to this question doesn't solve lack of ECC support for my i3-3250 and thus the need to switch to Xeon.

So you need to keep some things in context with this discussion over the Asus board.

Some brands have taken a CPU, put it in their motherboard, and if it works they call it "supported". They often do *not* validate every single little feature (like ECC). So even if you ask this question of Asus and specifically mention ECC, the drone on the other end of the phone may see the CPU is "supported" and feed you an answer so you'll hang up. This has been proven to happen before with MSI! They were caught red-handed a few years ago when someone was able to empirically prove that ECC did not function despite having emails that specifically stated that ECC is supported and works on the exact board he had proven otherwise on.

If you see me talk about ECC in detail, I use the words 'supported' and 'enabled' to specify whether someone claims it is supported and whether there is ample evidence that ECC is actually enabled and functioning. Many desktop boards will take non-ECC RAM and will work just fine. But we all know that the ECC RAM is not actually enabling ECC for that system. So you could call it "supported" because it does "work" (aka boots and plays your games, etc.), but it's not enabling a damn thing because neither the CPU nor motherboard actually use the ECC functions.

So regardless of what ASUS says, I don't trust them for a second. They are just starting to move into the server chipset market, so they are prone to errors, mistakes, and lack of attention to detail just like any other brand (I mentioned Asrock specifically with this topic earlier). The bottom line is that I have zero doubt that if Intel's docs say ECC isn't supported then Asus and any other manufacturer has no way of changing that. Intel protects their Xeon line and I don't buy for a second that there's some way that a motherboard manufacturer can buy a chipset and somehow change it so it will do what Intel doesn't want it to do.

So if you truly want ECC with that board, you'd better have a Xeon. Any motherboard manufacturer claiming otherwise is full of crap in my book. Who are you going to take at their word when Asus and Intel say the opposite thing? I'll trust Intel *long* before Asus.