Re: supervise not running

On Sun, 2008-09-14 at 22:42 +0930, Daryl Tester wrote:
> Larry Weldon wrote:
> 
> >> After the readproctitle, add an "echo `date` finished" to see if the
> >> script progressing past the svscan start, which it's not supposed to.
> >> Then reboot.
> 
> > did that. No additions to /tmp/svscanboot.log
> 
> So not even a "finished" message?  And yet svscan is no longer running?
> And svscanboot not running either?
> 
> That sounds incredibly borked if those three processes are dying with
> no further entries in the log file.
> 
> > Something further - as the system booted this lasst time I saw some
> > processes could not start and (I can't be sure - messages fly past
> > quickly) but it looked like it was because of messagebus.
> 
> Hmm, I can't think of any dependencies on messagebus that would cause
> an issue.
> 
> Later, in the next message:
> 
> >> Don't forget that init reads the inittab top to bottom. If your rc
> >> script (not the stuff in /etc/rc?.d but /etc/init.d/rc or whatever
> >> your system uses to actually run those scripts) doesn't complete, then
> >> init will never get to your svscanboot entry. See if rc.local or
> >> something like that never completes. The process list will show the
> >> existence of /etc/init.d/rc (or equivalent) if this is happening to
> >> you.
> 
> > That works. I put it right after;
> > id:5:initdefault:
> 
> OK, that's weird, because svscanboot was being at least initially
> started by init (you have a debug log to prove this), yet it appears
> to be getting terminated.  I wonder if something in one of the rc
> scripts is killing other processes?
> 
> You haven't modified the inittab order recently, have you?
> 

No - except in response to [email protected] and the latest on that is
that when I put svscanboot before the /etc/rc.d/rc ? scripts it runs OK.

When I put it after the rc scripts I found the haldaemon script did not
finish so I shut it off with ntsysv. The next boot I found the
following:

[root@tiger larry]# ps ax | grep rc
 2665 ?        Ss     0:00 /bin/bash /etc/rc.d/rc 5
 2783 ?        Sl     0:00 /sbin/prcsys S /etc/rc5.d/
 4679 ?        Ss     0:00 /usr/bin/imwheel -k
--rc /etc/X11/imwheel/imwheelrc.generic
 4720 ?        S      0:00 sh -c /etc/rc5.d//S92httpd start
> /tmp/init.BiXRq3 2>&1
 4731 ?        S      0:00 /bin/bash /etc/rc5.d//S92httpd start
 4858 ?        Ss     0:00 /usr/bin/lisa -c /etc/lisarc
 5947 pts/3    S+     0:00 grep --color rc

so now httpd doesn't finish.

> > env - PATH=$PATH svscan /service
> 
> My next recommendation was going to be to strace svscan, but if
> you're not getting the "finished" message then something (possibly
> external) is killing off processes that shouldn't be touched, and
> I suspect strace would be one of them.
> 
> 

well - [email protected] pointed out:
"If your rc script (not the stuff in /etc/rc?.d but /etc/init.d/rc or
whatever your system uses to actually run those scripts) doesn't
complete, then init will never get to your svscanboot entry."

which is what appears to be happening. I've started a virus scan.

I think you're right - strace will never be run.

Thanks,

Larry