07 March 2004

Once again, I must deal with RAM of Atijembar. After some kernel error reporting segmentation fault, I came into unstable condition yesterday and this morning. Examining log, I found that fatal errors happen when available free space was too low. In the kernel language, that is time to request new page (?) and reallocate unused space. Or I think that is time "high area" of RAM is used.

I reproduced this situation several times and, indeed, that is the problem.

One of Debian experts in #debian channel in Freenode suggested me to use memtest for testing my RAM. I tested using both memtest86 and memtest86+, but the results were somewhat strange. Even with new (again) replacement RAM, it reported some errors. So, which one can I trust?

The old 128MB one is still the best, no error was reported. I tried to install Debian with 128MB (enough, er?), but it was fail at the last stage of installation. Do not know why for a long enough time, and still suspected that it was RAM problem. On Friday, 5 March, after sholat Jumat and talking for while with Ismail Fahmi (unplanned meeting in Super de Boer), I came with idea this was maybe about partition problem which I used to ease (re)installation. The /var partition was mounted from old saved one, so I do not need to back-and-forth moving large size directory there.

I did installation with letting Debian creating its new /var partition, and... ah, successful! Soon after that, both new and old /var were synchronized, and Web server was up and run again.

For a month since now, I will test this RAM with various memory-minded operation to make it sure anything is in order.

No comments: