[plug] defunct process won't die
David Griffiths
griffith at environ.wa.gov.au
Wed Aug 9 13:27:33 WST 2000
>> Thanks for the help Steve.
>>
>> > 21989 ? DW 0:24 [netscape-commun]
>> > 21999 ? Z 0:00 [netscape-commun <defunct>]
>>
>> the ppid for 21999 is 21989
>> and ppid for 21989 is 1 as you suspected.
>>
>> However a
>> kill -HUP 1
>> does it's thing OK, but still leaves the two gremlin processess around.
>> (kill -s SIGHUP 1 complains about "s" being an unknown signal).
>>
>> Any further ideas very welcome!
>
>A couple of thoughts:
>
>pid 21989 is in a state of "uninterruptable sleep" with no pages in resident
>memory (DW) - ie it's not a zombie, which explains why a HUP to init
>didn't clean it up (pid 21999 *is* a zombie (Z) and killing 21989 should get
>rid of it) BUT
>
>Use top to tell you which process is in a runaway state (on my Mandrake
>system, when Netscape spaks, it's glibc20 which runs away with itself, hence
>my suggestion to grep the ps output for glibc20) and kill it. If that
>didn't clear the sleeping and zombie processes, then kill -HUP 1 should at
>this point. (kill -s is a BSD'ism, which is probably why it returned an error)
>
>Steve
Hmm... top doesn't throw up anything interesting that I can see (X, kvt and
top are the only processes taking up any cpu and they total about 2 %). The
glibc process isn't present at all in a ps and I can't see anything else
that looks like a runaway process, although my wmmon icon shows significant
(almost constant) activity - it must have some aggregate measure of load.
Might have to reboot and (sob) destroy my uptime :-(
re the kill -s, from the man page it looks like it should work, must be out
of date.
By the way the system is an upraded Red Hat 6.1.
cheers.
========================================================================
David Griffiths griffith at environ.wa.gov.au
Air Quality Management Branch
Department of Environmental Protection Phone: 61 8 9222 7151
Westralia Square Fax : 61 8 9321 5184
141 St Georges Tce
Perth 6000
Western Australia
========================================================================
More information about the plug
mailing list