[15:42] <keesj> Is there a way to listen to upstart events?
[15:44] <keesj> like the initctl events
[15:44] <sadmac2> keesj: not anymore
[15:44] <keesj> bugger 
[15:49] <keesj> I guess the "new" way of doing this would be to use dbus right?
[15:52] <sadmac2> keesj: doing what exactly?
[15:54] <keesj> I would like to supervise the events happening in upstart
[15:55] <keesj> mainly services stopping / crashing and such
[15:55] <sadmac2> yes. dbus should get you that.
[15:57] <keesj> with or without upstart hacking?
[15:57] <sadmac2> without
[15:59] <keesj> lets kick dbus-monitor then :P thanks
[16:11] <Keybuk> dbus won't get you that
[16:13] <sadmac2> Keybuk: we have signals for "service up" and "service down" don't we?
[16:14] <sadmac2> or at least you can watch for new/vanishing objects
[16:27] <Keybuk> right
[16:27] <Keybuk> there are signals for that kind of thing
[16:28] <Keybuk> just not generic events
[16:28] <sadmac2> Keybuk: yeah. so dbus gets him what he was fundimentally interested in seeing.
[16:28] <Keybuk> right
[16:29] <ion> keybuk: If you have a bit of free time, would you mind satisfying my curiosity by explaining how you’ve figured Upstart 0.6 would handle the fsck+mount thing? :-)
[16:29] <ion> s/explaining/describing/
[16:31] <Keybuk> oh, that's easy
[16:31] <Keybuk> you have a udev rule that parses fstab, and fills in the details
[16:31] <Keybuk> then you have an upstart job that matches those details
[16:31] <Keybuk> which calls mount
[16:31] <Keybuk> and then parses fstab again, and emits a can-mount event for anything underneath it
[16:31] <Keybuk> start on (block-device-added
[16:32] <Keybuk>   and can-mount $DEVNAME
[16:32] <Keybuk> exec mount $DEVNAME
[16:32] <Keybuk> post-stop exec what-can-we-mount $DEVNAME
[16:32] <Keybuk> etc.
[16:32] <ion> How will you handle blocking fsck instances that work on the same physical device?
[16:33] <Keybuk> you can do that with simple locking
[16:36] <ion> I have the prototype sh code that does some /sys traversing to get from /dev/some-lvm-device to some-md-device to [sda/sda1, sdb/sdb1] to [sda, sdb] somewhere. I could finish it; it could be used to get the list of names for the locks.
[16:39] <ion> flock(1) from util-linux could be useful here. It would need to be moved from /usr/bin to /bin, though.
[16:43] <Keybuk> the problem with the pure-upstart case though is how inefficient it is
[16:43] <Keybuk> you have to keep on re-parsing fstab
[16:44] <ion> Are you going to implement that for karmic nevertheless?
[16:44] <Keybuk> no
[16:45] <Keybuk> for karmic, I'm going with the "program that calls fsck, mount and swapon in the right order" approach
[16:50] <ion> Ok
[20:11] <wasabi> upstart have a 'user' option yet? 
[20:12] <sadmac2> wasabi: no.
[21:22] <wasabi> There a dh_* to install job files?
[21:24] <sadmac2> wasabi: I don't think I know what that is. Is it debian-related?
[21:24] <wasabi> Yeah
[21:24] <ion> 2008-10-15 17:46:28 < Keybuk> suihkulokki: I do not think dh_ anything is a good idea
[21:24] <ion> 2008-10-15 17:46:42 < Keybuk> what's wrong with just shipping the file as a conffile?
[21:25] <wasabi> Nothing. That's what I'm doing.
[21:25] <wasabi> Just wondering. I like dh_ stuff.