[03:53] <eXosypher> hellu
[03:53] <eXosypher> anyone alive? 
[04:01] <eXosypher> :(
[09:20] <vlt> Hello. I just upgraded from Ubuntu Edgy (which was a fresh install with upstart) to Feisty. Now the system only boots to runlevel "unknown". I always have to login and run `telinit 3` (sic!) to start all the daemons and the dm. `telinit 2` results in `runlevel` => "unknown", too. There's no /etc/inittab file and no "single" or similar argument in /proc/cmdline. What is wrong here?
[09:34] <ion_> Id suggest sending the question to the mailing list.
[09:40] <vlt> ion_: Ok, thank you.
[11:07] <docwhat> Hello!  Is logd supposed to be running all the time?
[11:07] <docwhat> This is ubuntu....
[11:09] <Keybuk> in edgy it does, in feisty it doesn't
[11:13] <docwhat> Ah.
[11:13] <docwhat> Where does it log output to?
[11:13] <docwhat> I'm sort of expecting behaviour like in supervise or runit.....
[11:13] <docwhat> I love the way the logging works there (ie, it uses stdout/stderr not magic logging functions).
[11:14] <Keybuk> /var/log/boot
[11:14] <Keybuk> logd captures stdout/stderr of running jobs
[11:15] <docwhat> Keybuk: Are you familiar with how supervise does it?
[11:16] <Keybuk> no
[11:17] <docwhat> It's similar.... There is a processes (one per daemon to be logged) that is kept always running (respawn).  This process just adds a timestamp to each line and writes it to a file with buffering options you set.  It's nice because your daemon doesn't need: sysklogd libraries, fancy logging routines, or behave differently depending on if you're testing it or running it production.
[11:17] <docwhat> logd with 'console logged' sounds similar.  Except for the running one per daemon.
[11:17] <Keybuk> though it does mean you end up with lots and lots and lots of processes
[11:17] <Keybuk> consuming lots and lots of resources
[11:18] <docwhat> The other neat thing is that the daemon can segfault without loosing logging output.  Equally, you can change the logging or role the log without restarting the daemon!
[11:18] <docwhat> The loging processes usually don't consume much resources.  A logging process is cheap compared to a daemon.  It just takes input and sends it to disk.
[11:19] <docwhat> You probably pay a small context switch cost vs. a single logging daemon (like sysklogd), but it is much much much simpler and easier to debug, configure and fix (and make secure!).
[11:19] <Keybuk> and much more difficult to restart when you do find a bug ;)
[11:20] <docwhat> Oh no, very easy.  As I said, the logger and the daemon are separate.  You can restart either/or without impacting the other.
[11:21] <Keybuk> yes, but you have to restart a couple of dozen logging daemons rather than just one
[11:21] <docwhat> Ah, I see what you mean.  Sure.  But the shell script for that would be: for i in /var/services/*/log; do sv restart $i ; done
[11:21] <docwhat> or something similar.
[11:22] <docwhat> actually, I think there is a command other than restart to do a cond-restart....
[11:22] <docwhat> (ie, restart if up)
[11:23] <docwhat> *shrug* another alternative is just make the init.d for the daemon include the logger there, but I figured since upstart already had a runit/supervise feel to it that this would be a nice addition.
[11:23] <Keybuk> it's pretty much what we already do, no?
[11:23] <docwhat> Thanks for the chat, Keybuk!  I have to go.   Have a good day. :-)
[11:23] <docwhat> Yup.