Hi all

 

After running 4.2.1 since January without issue I did a number of things recently. I upgraded to 4.3.3. I upgraded my parity drive to a 1 TB WD green drive. I took my former parity drive and made it my cache drive. I also changed cases and moved everything over to the new Centurion 590.

Anyway, within a week of upgrading my stuff, I've seen thre kernel panics. Obviously I can't telnet into the machine to get at the syslog, but I did take down the message that I got. I think it's been the same each time:

 

CPU 0: Machine Check Exception: 0000000000000004

Bank 0: b200004000000800

Bank 5: b200121014040400

Kernel panic – not syncing: CPU context corrupt

 

Any help appreciated. Thanks,

 

Joe

Looks like memory.  Not seated properly, bad, wrong voltage, whatever.

 

 

Bill

Thanks Bill. I wondered whether the reference to 'Banks' meant the RAM. Perhaps I zapped it when moving the components over to the new case. I'll try reseating and see how it goes.

Thanks Bill. I wondered whether the reference to 'Banks' meant the RAM. Perhaps I zapped it when moving the components over to the new case. I'll try reseating and see how it goes.

A lot of today's memory need to have the voltage set higher than standard for a motherboard.  Recommend you run the memory test for a few cycles once you re-seat the memory.

 

Joe L.

I've noticed that. However, the RAM was fine for a good 7 months while I was with 4.2.1, so unless the requirements for 4.3.3 are different, I think the voltage should be okay. I may give it a little tweak if memtest requires it. Thanks for the tip.

More info:

 

This seems to happen when I stop the array. I've shut down the server since I'm at work now, but when I start it up again, I'll do the memtest and see what gives.

3 passes of memtest with no errors.

 

So, other ideas included reverting back to my Netgear ethernet card to see whether the driver of my Broadcom could be causing trouble. Unfortunately, the Broadcom NIC is a PCIe one and the Netgear was a PCI one, which was in the only spare slot; that slot is now filled with a 4 port SATA card. I could try using the onboard nVidia gigabit interface.

 

Any other thoughts?

I could try using the onboard nVidia gigabit interface.

 

Onboard nVidia gigabit usually works fine, drivers are included in unRAID.

I'm leaning more and more towards my PSU not having enough juice. It's a new Antec 650W unit. I have 9 HDs and 8 fans (120 mm) in total, with drive cages and case fans.

On startup, all the fans spin up, but then 2 of them stop once the HDs spin up.

From what I've seen. the kernel panic usually happens on a 'stop array' command being sent. From another thread, I'm assuming that this command causes the drives to spin up so that they can be unmounted? I'm wondering, if so, whether the power spike is fatally taking away juice form the RAM/CPU?

 

In any case, I'm goign to upgrade to an 850W PSU very shortly. Perhaps todays beefy PSUs are designed mostly for power hungry video cards, rather than arrays of drives all spinning up together?

Probably should make this a sticky -

 

Most high-wattage PSU's split their output into multiple rails, where each rail is current-limited typically to 12A, though some are higher.  The point is that usually only 1 rail is allocated to peripheral devices, i.e., your hard drives.  So even if you used PSU rated at 36A, if it has multiple rails, then all your hard drives are going to be on a single common 12A rail.

 

There are single-rail PSU's out there, for example PC Power & Cooling, which is what we use in our servers.

Probably should make this a sticky -

 

Most high-wattage PSU's split their output into multiple rails, where each rail is current-limited typically to 12A, though some are higher.  The point is that usually only 1 rail is allocated to peripheral devices, i.e., your hard drives.  So even if you used PSU rated at 36A, if it has multiple rails, then all your hard drives are going to be on a single common 12A rail.

 

There are single-rail PSU's out there, for example PC Power & Cooling, which is what we use in our servers.

 

Thanks for the info. My current PSU is an Antec Trio 650 W one, with three 19A rails.

The one I'm upgrading to is an Antec Quattro 850 W with four 18A rails. Does that mean for the purposes of an unRAID server, I'm actually downgrading?

I had this exact same issue.  Stepped up from 12 to 14 drives and carrying out the parity check taxed it too much and I got lots of errors.  Swapped out to a new 850w PSU and the thing has been sweet for months.

Probably should make this a sticky -

 

Most high-wattage PSU's split their output into multiple rails, where each rail is current-limited typically to 12A, though some are higher.  The point is that usually only 1 rail is allocated to peripheral devices, i.e., your hard drives.  So even if you used PSU rated at 36A, if it has multiple rails, then all your hard drives are going to be on a single common 12A rail.

 

There are single-rail PSU's out there, for example PC Power & Cooling, which is what we use in our servers.

 

Thanks for the info. My current PSU is an Antec Trio 650 W one, with three 19A rails.

The one I'm upgrading to is an Antec Quattro 850 W with four 18A rails. Does that mean for the purposes of an unRAID server, I'm actually downgrading?

 

Could very well be.  It depends on how the connectors are divided among the rails.  Looking at the Antec website for the PSU, there doesn't seem to be anything that defines this.  Perhaps you might try emailing their tech support.

Thanks Flambot and Limetech. I'll try removing my cache drive. I don't even know how to use it, so I'll see if that spares me a few problems.

Also, yes during a parity check, one of my drives just failed (red ball icon). Bummer, as I finally got the temperatures of the drives down to something acceptable. 

I'm guessing that pulling tha cache drive and restarting will be just fine.

 

Here's an update.

 

I think it's very likely that the problem was a PSU related one. I haven't changed my PSU yet, but because of a heat problem, I rewired my whole server, moved some disks around etc. I reconnected fans and drives in my Centurion 590 case so that the two moex cables had a more even mix of hard drives and fans to power. I replaced a Seagate drive with a 1 TB green power WD drive as well. Now I have 9 drives and 8 120 mm fans in the rig and it's working great. The rewiring also helped the airflow and it's running much cooler than before too. The good news is that an even beefier PSU is on its way and I think I'll be okay with it.

 

I also noticed recently that J River Media Center (my HTPC's front end) was far more responsive when I accessed my server some time after the drives had spun down. I was really happy about this, as before it used to be slow to get my browsing of media files going. I assumed it was because of the frequent updates made by the JRMC staff; now I see it's much more likely that it was because of the simultaneous, rather than staggered, spin up of 4.3.3 (my previous version of unRAID was 4.2.1).

 

Thanks for your help. I've been making computers for years, but never with so many drives/fans, and so I've learned a few things.

The cooling could be helping as much as the more balanced load on the PSU as the efficiency of a PSU drops significantly with increased temps.

 

Good to hear everything is stable.

 

 

Bill

Looks like I spoke too sonn.

All was okay until I added a 9th drive, and I think that last drive pushes the system over the edge. It starts up okay, but spinning up or stopping the array from the web interface causes a kernel panic. The newest drive isn't a part of any of my user shares at the moment, so I'll just remove it for the time being. When my new PSU arrives, I hope I'll be able to add to my array.

Funnily enough a new problem has cropped up now - renaming any file on my unRAID box from my desktop PC causes the desktop PC to reboot. I can rename anything I want from my HTPC.

I was able to duplicate the kernel panic from my HTPC, so it seems to be a hard/software issue on the unRAID box itself.