|
Here are two questions I've had since being an admin that I've not seen answered yet. Why can't I forcibly remove 'z' state (zombie) processes from the process table to reclaim PIDs/space, and why can't I forcibly kill a process stuck in 'd' state (disk wait)? Our machine used to collect about 5 zombies per day for some reason, and we ran out of proc table space eventually on it. Also, since we ran on NFS, sometimes we REALLY needed to kill processes that were uselessly stuck in disk wait. I never felt like it made sense to have to wait 10 minutes to get my terminal back because the net dropped while I was doing 'ls'.
|
#
¿
Mar 24, 2007 18:32
|
|
|
|
| # ¿ Apr 26, 2024 08:49 |
|
|
dfn_doe posted:What kernel version are you running? The kernel should go through periodically and clear out zombied procs, at least in newer versions. I have no clue about the diskwait question though. Well, it was 2.2.19 at the time. Some scientist started a fork bomb by accident that would have been fine except for filling the process table (65535 entires, I think?) with zombies. I was told letting the admins remove zombies was a Bad Idea, but it was never explained why.
|
|
#
¿
Mar 24, 2007 18:42
|
|
|
Wedge of Lime posted:I am not entirely sure if this is correct for Linux, however under Solaris if you imagine that a program which opens a file over nfs must ask the kernel to do it, as it is a privileged operation. Oh, wow, that's the first time someone's ever explained the technical side of it, which is what i've been wondering. And yes, the obvious solution has always been to fix NFS. Given the technical difficulties of using NFS between Linux 2.2.17 and Solaris 2.7 pre-NFS patching, this is beyond the point. Thank you.
|
|
#
¿
Mar 28, 2007 07:31
|
|