PS/Top broken - /proc entry bad

I had a box where ps and top quit working after hundreds of days uptime. After doing an strace ps I found that one directory in /proc was hanging it up, a directory named a 5 digit number which I believe was associtated with a process of the same name.

I tried doing a kill -9 on the process, it returned fine but the process was still there. Reboot hung my session, too, I had to use reboot -f to get the machine healthy again.

Is there any way to "fix" /proc other than what I did? I suppose maybe going into a lower init level and then back to 3 may have worked. It's a remote machine, though, so reboot was at the time seemed like a better solution.

Any comments/suggestions on what to do in this situation?

Thanks
Matt

Use sysreq-t to get a backtrace of the processes. Most likely one of them hung while still holding the mm semaphore, thereby preventing ps and top from proceeding. Check your log for oopsen.

-ben

"Do you seek knowledge in time travel?"