[RFC] Logging events and jobs in Upstart

[Scott James Remnant]

Remind me on Monday to post a script I use.

It parses the kmsg log, and the output from D-Bus initctl calls, and
generates a giant dotty chart of the boot sequence - not just based on
configuration but on what actually caused what

On Fri, Jun 17, 2011 at 1:09 PM, James Hunt <james.hunt at canonical.com>wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi All,
>
> I'd like to get some discussion going about the logging of events and jobs
> in Upstart.
>
> Scott has done some great work recently wrt the kmsg change [1] such that
> you can now see everything
> from the "startup" event onwards if you boot with --debug/--verbose and
> configure your syslog daemon
> appropriately.
>
> The ability to have a log of all Upstart events is invaluable, particularly
> for servers which have
> been up for aeons as it helps users and admins learn more about their own
> particular set of event flows.
>
> I'm wondering therefore if it is worth taking the logging a step further by
> considering one of the
> following (or any other ideas readers have of course):
>
> 1) Making Upstart default to logging event/job state messages.
>
>   We might need to change some of the formatting a little to make it more
> palatable to admins, but
> the work is essentially already done.
>
> 2) Have Upstart store a log of event+job details *internally*.
>
> I'm not suggesting we necessarily do this, but I think it's an interesting
> idea so I'd like to share
> it...
>
> The thought is that we introduce a new configure option (default disabled)
> which if enabled would
> make upstart store internally the name/env/exit status? (and maybe some
> sort of timestamp) for every
> event and job it runs.
>
> The problem with this is:
>
> - - the data isn't viewable anywhere.
> - - assuming a linear log [2], more and more memory is being consumed (a
> bad thing for init generally)
>
> The solution to this is 2 new initctl commands:
>
>        initctl show-internal-log
>
> This command would write the internal log to stdout. However, if you ran it
> like this...
>
>        initctl clear-internal-log
>
> This would make init free its internal log.
>
> Now, as long as clear-internal-log is called at some "reasonable" point in
> the boot, we have the
> ability to see exactly what events were emitted and what jobs were run, the
> relative order and we
> stop init gobbling up ever increasing amounts of memory.
>
> Comments?
>
> Kind regards,
>
> James.
>
> [1] -
> https://lists.ubuntu.com/archives/upstart-devel/2011-February/001409.html
> [2] - we could of course consider a circular log.
>
> - --
> James Hunt
> ____________________________________
> Ubuntu Foundations Team, Canonical.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.11 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>
> iEYEARECAAYFAk37tIkACgkQYBWEaHcQG9dkSgCfQMkoUw2VjIRjVXAH7xJdXUAt
> udcAnjNtyX0SNY0N2sh6hRU8U/2X1k7g
> =vk6z
> -----END PGP SIGNATURE-----
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/upstart-devel/attachments/20110617/34da5c94/attachment.html>