Re: LYNX-DEV Lagging Processes

[Jonathan Sergent]

[longish]

Daniel Spillers writes:
 ] i'll check those ... i've seen both the ps and top feeds, and it just
 ] looks like the processes are sitting there (with no one connected to them,
 ] re: the who) doing who knows what. the more there are, the slower the
 ] machine flies.


Do the processes have ttys? (in the 'ps' output, does the TTY column
say "?" or does it list a tty on which nobody is logged in?)  Mainly
curious.  If they do have ttys, is the telnetd or rlogind which 
connected the user still running?  Is the users' shell still running?
Does lynx or the shell list "1" in the PPID field ('ps alx' to see it)?
It sounds like you're saying that PS shows that they do have controlling
ttys but there's nothing else (rlogind, telnetd, etc.) on said tty.

I suppose there could be a problem with lynx signal handling, but I
doubt that as we haven't seen lynx do this before; the signal 
handling code is quite complex, though, if I recall.

I have seen some weirdness with signal handling and RHL4.1,
but never spun processes.  I use lynx quite frequently on several
4.[01] boxes, and it's quite reliable for me.  And although I see
That Other Browser spinning quite frequently on the large multiuser 
(Solaris) systems here, I've never seen lynx do it.

That would suggest it might be something strange with your network,
or the way your users are logging in [as Henry suggested].  If you
can't get people to log out properly, your best bet is probably to
just kill them.  You could probably write a script (and periodically
run it via cron) to detect and fix them.  Someone more familiar with
the signal handling involved might be able to devise a workaround
for Lynx, although I'm not sure.


-- 
Jonathan Sergent / address@hidden / address@hidden
;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;