Hardware Error in syslog

kfritz · 06-17-2014, 10:06 PM

I've not seen output like this before:

Code:

Jun 17 19:52:01 hex kernel: [137282.075199] [Hardware Error]: MC1 Error: Parity error during data load.
Jun 17 19:52:01 hex kernel: [137282.075206] [Hardware Error]: Error Status: Corrected error, no action required.
Jun 17 19:52:01 hex kernel: [137282.075209] [Hardware Error]: CPU:2 (10:a:0) MC1_STATUS[-|CE|-|-|AddrV]: 0x9400000000000151
Jun 17 19:52:01 hex kernel: [137282.075211] [Hardware Error]: MC1_ADDR: 0x00007f1827e21260
Jun 17 19:52:01 hex kernel: [137282.075212] [Hardware Error]: cache level: L1, tx: INSN, mem-tx: IRD

Anyone have a clue what's failing here? I was running a few VM's pretty hard during this time...

ReaperX7 · 06-18-2014, 01:51 AM

Looks like a CPU core timing error. What kind of CPU are you using?

kfritz · 06-18-2014, 09:41 AM

AMD Phenom(tm) II X6 1100T Processor

ReaperX7 · 06-18-2014, 03:39 PM

Are you using single core per VM or multi-core?

kfritz · 06-18-2014, 06:52 PM

All my VM's are running single CPU, but I'm not setting any affinity, if that's what you're asking. I was running some upgrades on 4 VM's simultaneously, and had another running that was fairly idle at the time.

ReaperX7 · 06-18-2014, 07:00 PM

What level of % of CPU resource per core do you allow for each VM?

volkerdi · 06-19-2014, 12:54 AM

Quote:

Originally Posted by kfritz

AMD Phenom(tm) II X6 1100T Processor

I have that CPU, too. It's thrown that error a few times when the heatsink needed to be cleaned.

anscal · 06-19-2014, 04:46 AM

Are you using ECC RAM modules? It may be that a memory error has been corrected through ECC.

kfritz · 06-19-2014, 10:21 AM

Quote:

What level of % of CPU resource per core do you allow for each VM?

I'm just running qemu-kvm (qemu-system-x86_64) with some pretty generic parameters:

Code:

qemu-kvm \
    -net vde,vlan=0 -net nic,vlan=0,macaddr=DE:AD:BE:EF:00:0C \
    -m 1G \
    -drive file=disk.bin,index=0,cache=writeback \
    -enable-kvm \
    &

Quote:

I have that CPU, too. It's thrown that error a few times when the heatsink needed to be cleaned.

Hmm, I'll have to take a look this weekend. Damn cats!

Quote:

Are you using ECC RAM modules?

Yes. I actually forgot that I spent the extra $$ on that.

metaschima · 06-19-2014, 10:57 AM

Looks like a CPU L1 cache error, but it was corrected. As long as everything still works without crashing it's ok. Cleaning the heat sink and other fans is a good idea.

Even if one of the cores fails at the level of the L1 cache, you can still run the system with fewer cores with some kernel boot options to disable the failed core. So far this is not needed.