Adam Hawes wrote: >> Any idea how to diagnose the problem? > > Turn on as much verbose logging as you can and see what you get. Do you > get any kernel crash messages on the console when the machine hangs or > does it just hang up and die?
No, there are no kernel messages and nothing in the log files. It just hangs up and die. > Memory can be stressed by hefty compiles. Building the kernel is a good > test. > Does a kernel compile succeed? How about if you use make -j 4 to run 4 > tasks in parallel and use up more RAM? That also stresses the CPU. You > could boot one of those *cough* Linux live CDs with a memory tester on it and > run a memory test overnight to look for failures but that's not reliable in > many > border cases. > > Are the disks OK? Smart can be used to check for some kinds of errors. > Look in the logs for disk access failures. The disks are OK, SMART is enabled. > Finally, if you think it's network related unplug the network and stress test > the machine. Try a different NIC if you have one lying about. I will try with compiling to stress the memory and CPU. Regards, Mitja