Hi all,

I have a 'Rackable Systems' server with an HDAMA mobo - Dual CPU Opteron 250 2.4GHz with 2x memory modules per CPU. It seems to run fine, but it hangs every hour or so!

I am running Ubuntu 64-bot 10.10, which is currently a beta release so I haven't discounted that as the problem yet, but suspect it unlikely. However, I am downloading 10.04 as I type...

dmesg spits out lots of awful messages like these:

[ 1314.920127] EDAC MC1: CE - no information available: amd64_edacError Overflow
[ 1315.920047] Northbridge Error, node 0, core: 0
[ 1315.920060] ECC/ChipKill ECC error.
[ 1315.920066] EDAC amd64 MC0: CE ERROR_ADDRESS= 0x1484410
[ 1315.920082] EDAC MC0: CE page 0x1484, offset 0x410, grain 0, syndrome 0x11c1, row 0, channel 0, label "": amd64_edac

(there are variations on the node, core, address, offset an syndrome etc.)

I have tried swapping CPUs over and running with only CPU.
I have also swapped all the memory around in almost every permutation.

Another worrying symptom is that when I run memtest86+ from a boot disk, it shows zero errors up until the point where the server turns itself off without warning - it hasn't yet completed the test...

If anyone could shed some light on this, I would be grateful. Perhaps I've bought a dodgy second-hand computer, so steep learning curve. But it bugs me not knowing what the root cause is...

Thanks,

Ben

So memtest never finishes ? I would try removing a RAM stick and seeing if it still crashes, if it does, put it back and take out another one.

That is the address of the bad sector of RAM. Different hardware vendors can use this to tell you which DIMM is bad. For example on our old Sun v20z machines this would be DIMM0 on CPU0. However on our Sun x4000 series machines it does not map to the same DIMM.