Kernel panic - bad memory?

[tristanlee85]

I have been trying to import and 2gb SQL file and during the import, the system locks up. I finally hooked up a monitor to the server and noticed that I get the following message:

ACPU 0: Machine Check Exception: 0000000000000004
Bank 0: b20000001020080f
Kernel panic - not syncing: CPU context corrupt

I have also added the "nomce" argument to the grub config to see if I can run the import again without it failing. If it doesn't fail this time, could it be memory, CPU, or what?

[tristanlee85]

Well, nomce didn't help things. It still locked up.

[Rubberman]

Well, the message is telling you that memory is going south for the winter. It is even telling which bank of memory is failing and which chip. This is one reason why I ALWAYS install ECC RAM in my workstations. It will give me an error, but it will normally be able to correct it and continue to run, allowing me a bit of time to order and install a new part. In fact, that has happened to me in the past and the ECC saved my bacon in a big way. In any case, I positively will not purchase a system or motherboard that does not support fully buffered ECC RAM.

[tristanlee85]

Thanks. I wasn't sure if it was memory-related because the search results I pulled up had mixed opinions.

[Rubberman]

A database search will be (very) unlikely to cause a memory failure like this. However, importing a large SQL file will use a lot of memory, so it is more likely to hit the bad RAM. Any mission critical system I will always install with ECC RAM, even if it is more expensive.