I've been having a recurring problem on my Linode box where every month or so all four CPU cores spike to 100% and the box is essentially dead. I had setup top -b in crontab but it wasn't able to get cycles to run -- the top output stops exactly when the CPU spikes. I also checked sar for that day and it stops as well. As a next step, I've started top in the background with nice:

nohup nice -n -20 top -b -c -d 60 > /tmp/top.out 2>&1 &

Any other ideas on how to capture which processes are causing this? I see nothing in /var/log/messages

Thanks

Just a thought - if it's happening monthly perhaps it's some cron task that's actually causing the issue.