[01:25] <Keybuk> wasabi: so let me try and explain :)
[01:25] <Keybuk> you have a job for a web server
[01:25] <Keybuk> in this job, you define the configuration for the web server itself; what needs to be run, how to prepare the machine and clean up afterwards, etc.
[01:25] <Keybuk> any chroot/chdir/resource limits/etc.
[01:25] <Keybuk> this is the "job"
[01:25] <Keybuk> you also define the machine state in which the web server should be running
[01:26] <Keybuk> these are events and pairs of cancelling events
[01:26] <Keybuk> or references to other job/states
[01:26] <Keybuk> e.g. the web server is running while there is at least one non-local network interface up and the filesystem has been assembled
[01:26] <Keybuk> this is "the state in which the web server can be run"
[01:27] <Keybuk> and the web server job state is nominally coupled to this state
[01:27] <Keybuk> any changes to this state change the goal of the web server job
[01:27] <Keybuk> however this is not "the state in which the web server IS running"
[01:27] <Keybuk> because the web server could fail to start, and stop again, without the underlying state being changed
[01:27] <Keybuk> or the web server could be started and stopped manually
[01:27] <Keybuk> or the web server could be a member of a disabled profile
[01:27] <Keybuk> etc.
[01:28] <Keybuk> jobs that reference "when a web server is running" are NOT referencing the state in which the server could be running, but the state in which it IS running
[01:28] <Keybuk> so there's a second state here
[01:28] <Keybuk> one state is defined in the job configuration and is tied to the goal of the job
[01:28] <Keybuk> the other state is implicit and is tied to the state of the job
[01:29] <Keybuk> so "with apache" means the state of the apache job
[01:29] <Keybuk> not the state in which the apache job can be running
[02:22] <wasabi> i see
[02:33] <Keybuk> does that make sense?
[04:05] <wasabi> does this # have logs?
[04:06] <Keybuk> yeah
[04:06] <Keybuk> people.ubuntu.com/~fabbione/irclogs/upstart-devel-....
[04:06] <Keybuk> err
[04:06] <Keybuk> upstart-... :p
[04:26] <Keybuk> http://upstart.ubuntu.com/wiki/States
[06:13] <AlexExtreme> heh, the implementation of profiles i did before was purely lazy
[06:14] <Keybuk> hmm?
[06:15] <AlexExtreme> well, it had absolutely no support for changing profile at runtime, in fact it was quite easy to do that
[06:15] <AlexExtreme> it used multiple variables to store info about the current profile rather than a struct
[06:16] <AlexExtreme> so i've had to do a fair bit of work to get it working with the new config code, rather than changing a few lines
[06:16] <AlexExtreme> laziness != good ;)
[06:27] <Keybuk> oops, sorry
[06:36] <AlexExtreme> \o/
[06:36] <AlexExtreme> parse_profile.c written
[06:37] <Keybuk> never read code you wrote when you were 18
[06:38] <Keybuk> it's very disturbing
[06:38] <AlexExtreme> heh
[06:38] <AlexExtreme> i have code here that i wrote when I was 9
[06:38] <AlexExtreme> it's *very* disturbing ;)
[06:38] <AlexExtreme> not to mention that it's written in VB
[06:39] <ion_> Anything VB is disturbing.
[06:39] <AlexExtreme> true
[06:40] <ion_> Speaking of disturbing and VB, http://johan.kiviniemi.name/blag/2006/12/26/excellence/ (in you havent been spammed with this URL yet).
[06:41] <AlexExtreme> interesting
[06:47] <wasabi> blah. this new server will not run mdrun and vgchange properly in the initramfs
[06:47] <wasabi> these packages are constantly pissing me off heh
[06:48] <Keybuk> wasabi: which release?
[06:48] <wasabi> feisty
[06:48] <Keybuk> heh, yeah, that was pretty bustef
[06:48] <wasabi> My desktop on gutsy is busted too.
[06:48] <Keybuk> yeah, gutsy is more busted righ tnow
[06:48] <wasabi> ... my desktop is weird. It looks like udev spawns a ton of mdadm processes.
[06:48] <wasabi> infinitly.
[06:48] <Keybuk> evms installed?
[06:49] <wasabi> Nope, I've removed EVMS from teh picture.
[06:49] <Keybuk> oh right
[06:49] <Keybuk> dunno then
[06:49] <wasabi> You win.=)
[06:49] <Keybuk> I do?
[06:49] <wasabi> Yup. No more evms for me. heh
[06:50] <Keybuk> lol
[06:50] <Keybuk> "make wasabi use ext3" <g>
[06:50] <wasabi> Are you still familiar with the local-top/mdadm|run scripts, or is somebody else doing them now?
[06:50] <wasabi> Oh, you won that too.
[06:50] <Keybuk> I'm trying to understand what they do/are for
[06:50] <wasabi> I just reformatted to ext3.
[06:51] <wasabi> Mostly because half of my files because mysteriously truncated.
[06:51] <Keybuk> heh, it's interesting how most people abandon XFS after an "event"
[06:51] <Keybuk> mine was when a bunch of files had \0 through them for no readily apparent reason
[06:52] <wasabi> ldconfig: file /usr/lib/libwine.so.1 is truncated
[06:52] <wasabi> at least ldconfig is very clear about the problem. heh
[06:52] <wasabi> RUN+="watershed -i udev-mdadm /scripts/local-top/mdadm from-udev"
[06:52] <ion_> keybuk: Ditto.
[06:52] <Keybuk> yeah
[06:52] <wasabi> That's an odd one. What's the 'udev-mdadm' in there refer to?
[06:52] <Keybuk> I don't understand why ian runs that
[06:53] <Keybuk> since in the real filesystem, we just run mdadm itself
[06:53] <wasabi> You mean mdrun?
[06:53] <Keybuk> err, mdadm I think
[06:53] <Keybuk> I don't really understand what the difference between the two mdadm related initramfs scripts is
[06:53] <Keybuk> and what the difference between mdrun and mdadm is
[06:53] <Keybuk> once I understand, mdadm shall be fixed
[06:53] <wasabi> Looks like mdrun does auto detection. It's a single shot line that probes all drives.
[06:54] <wasabi> Probably not what you need with udev feeding events.
[06:56] <wasabi> Hmm. Yeah, looks like mdadm has the work mdrun does duplicated... just in a harder fashion.
[06:56] <wasabi> ... builds a config file on the fly from --scan, and then uses that.
[06:59] <Keybuk> interesting
[06:59] <Keybuk> neither seems right though, since we need to run something repeatedly until it sticks?
[06:59] <wasabi> Yeah. I think what mdrun does is technically rigth... probes all devices.
[06:59] <wasabi> Just watershed it on any device event.
[07:00] <wasabi> Oh except mdrun is deprecated.
[07:01] <wasabi> well that cinches that. *deletes the file*
[07:04] <Keybuk> lol
[07:10] <Keybuk> welcome to my world
[07:10] <Keybuk> "ah, this looks exactly right ... wait, what'd'ya mean deprecated?"
[07:11] <ion_> :-)
[07:16] <AlexExtreme> :)
[07:18] <Keybuk> you rained yesterday, so flying was cancelled
[07:18] <Keybuk> you're raining again today
[07:18] <Keybuk> you better not rain on Saturday
[07:22] <AlexExtreme> heh. sometimes I wonder whether this really is summer...
[07:22] <AlexExtreme> i think we had summer here in april ;)
[07:23] <Keybuk> yeah
[07:25] <ion_> Not very summer-ish here either, but thats to be expected, this is Finland after all. :-)
[07:59] <AlexExtreme> well, that's a start
[07:59] <AlexExtreme> it segfaulted
[07:59] <AlexExtreme> :p
[08:07] <AlexExtreme> bah
[08:07] <wasabi> honest question. how do you know what is "supported" and not "supported" on ubuntu. md, for instance. How can it be that it is knowingly busted in feisty?
[08:07] <wasabi> Heh.
[08:07] <AlexExtreme> it helps to call nih_list_init on a NihList before using it
[08:08] <wasabi> In fact who specifically should be tarred and feathered for this?
[08:09] <AlexExtreme> i'm not sure about md in feisty, i used it once it it was broken, kinda.
[08:09] <AlexExtreme> can't remember what happened exactly
[08:10] <AlexExtreme> brb, testing again
[08:11] <wasabi> Or perhaps it's my fault for thinking feisty would work at all the same things dapper did. Perhaps the lack of the LTS.
[08:12] <AlexExtreme> neat
[08:12] <AlexExtreme> it worked
[08:16] <wasabi> hmm. wonder if mdadm and vgchange could somehow deadlock on each other.