Re: supervise not running

Larry Weldon wrote:

>>>> Run "svscan /service" from the command line as root to see if any
>>>> errors are being directly reported.

>>> OK. That works but it doesn't show anything which was not already
>>> detailed in my first post.

>> You didn't report that readproctitle had reported anything.

> readproctitle was not in the process list at all.

So when you ran "svscanboot &" from the command line, readproctitle
was not running at all?  I'm already aware that readproctitle wasn't
running after boot, but it should have run when you started it
manually.

> the path is /command/svscanboot

I meant path as in PATH, which lists multiple directories (in case
of multiple installed copies).

> the command line in svscanboot which is supposed to start it all...
> /command/svc -dx /service/* /service/*/log
> 
> I thought svscan was supposed to start it...

It does - svc -dx terminates any already running supervised instances
in case svscanboot is being rerun due to failure.  You see in the log
below that the terminate failed because supervise wasn't running, as
it should be.

> contents of /tmp/svscanboot.log:
...
> svc: warning: unable to control /service/dnscache: supervise not running

As expected.

> + env -
> PATH=/command:/usr/local/bin:/usr/local/sbin:/bin:/sbin:/usr/bin:/usr/sbin:/usr/X11R6/bin svscan /service
> + env -
> PATH=/command:/usr/local/bin:/usr/local/sbin:/bin:/sbin:/usr/bin:/usr/sbin:/usr/X11R6/bin readproctitle service errors:

OK, this is how I would expect it to end if it was running successfully.
If there is nothing in the log file after that, and still no svscan running,
then there is multiple weirdness here because I would expect init to rerun
the svscanboot script (it was set to respawn) and thus append to the log
multiple times (check that it definitely was "exec >>" at the beginning
of the script so the log file is appended to and not overwritten).

A couple more things to check.  See if the svscanboot script is still
running, and modify the svscanboot script further (hey, you've come
this far) to remove the "2>&1" after the "svscan /service" (so stderr
from svscan should be capture by the log script, and not readproctitle).
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.

As a second (third) stab, you might also want to modify the script
to run svscan on its own (i.e. as you did from the command line),
to wit:

env - PATH=$PATH svscan /service

This will cut readproctitle out of the loop completely, which my gut
feeling is telling me that's its terminating voluntarily because svscan
is terminating (the above log implies that it at least both ran) and
taking useful log info with it.  But that's just edumacated guesswork
at the moment.


-- 
Regards,
  Daryl Tester

"I'm beginning to wonder about Cap'n Snooze's military credentials."