Hi All. this gets a bit long winded, but please bear with me.

The short version is what is the best (most intensive) way to test
hardware integrity, as well as isolate the exact problem?

yesterday I upgraded my box. I bought a new case/mb/cpu/ram and used all
the other existing components.

It's a AMD K7 500 w/ 256MB ram btw :)

I am currently running a kernel compile in a while /bin/true; do loop and
using the box at the same time as a "burn in test". It's been running like
this since yesterday sometime. I'm quite happy.

Oh so I should ahve mentioned that the old box used to lock up pretty   
frequently. I used to attribute it to the fact that I used to have scsi 
warnings on bootup, and felt like the drives were running too hot. Hence
the upgrade to a full tower case with three case fans.

Now, the real reason for this post is that I took the old case/mb/cpu/ram
combo (K6 II 450/ 256mb), stuck in a new ide CDROM, a new ide HD, and two
new netgear nics. My intention is to make this a server for my mom who
lives a 1000 miles away. It needs to be reliable.

Well the old mb/cpu/ram combo locked up on me twice already. I ran the
kernel compile loop on this box as well. It made it overnight, but locked
up sometime this morning. I ran it again, and this time it locked up in
just under 3 hours. I am currently running it again with one of the 128MB
dimms removed.

Nothing in the logs when this occurs.

So, I'd really like to figure out exactly which component needs
replacing. I'm guessing it's likely in the order of ram, then cpu, then
mb? But, I'd also like to be darn sure I test these properly. This really
needs to be reliable. The last box I setup for my mom turned out to have a
memory problem - only found out once it was installed and I was trying to
build something on it. THen lightning took out the modems so I could no
longer remotely admin it (but that's a whole other problem).

I have also just kicked off a program called memtester
http://www.qcc.sk.ca/~charlesc/software/memtester in the hopes of catching
a memory problem.

thanks
charles


-- 
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.

Reply via email to