/srv/irclogs.ubuntu.com/2010/08/03/#upstart.txt

=== Md_ is now known as Md
ionkeybuk: A user thinks mountall’s ionice code doesn’t actually ionice fscks properly on his system. That diagnosis hasn’t been verified by me yet, but could my code be wrong due to me misunderstanding process groups? After fork and before exec(fsck), i do setpgid (0, 0) (the fsck is now supposed to be the leader of a process group with the same id as the pid, right?) and the priority setting is done with ioprio_set (IOPRIO_WHO_PGRP, fsck_pid, ...)16:02
ionI did that in case some fsck implementation forks children which do the havy IO.16:03
ionheavy16:03
Keybukright16:04
ionSaid user ripping all of upstart, mountall and plymouth out of his system as a reaction doesn’t help with debugging the issue he had. :-D16:08
Steveehello Keybuk, is there any way the see the current development state of the next upstart version ? 16:21
Keybuknope16:22
Steveeat launchpad i only can see that "trunk" but there aren't any changes16:22
Steveeoh, thats bad, what will be the next version ? 0.6.7 or any other release number ?16:23
Keybuk0.10.016:23
Steveewow, great16:23
Steveewhat do you think, when it will be released ?16:24
Keybukwhen it's ready16:24
Keybukit's a development project at the moment16:25
Keybuksome very big changes under the hood16:25
Steveeoh, i see16:25
Keybukso I'm still ripping things apart and experimenting16:25
Keybukthus it's not in any state I'd be happy to commit to revision control16:25
Steveei see, so you aren't able to guess any release date or period16:26
Keybuknope, not yet16:27
Steveebut good to hear, that development still goes on16:27
Keybuk0.6 is stable and useful16:27
Steveeyes of course, we are using it in the current development version of our distribution16:28
Steveeand it's great - thanks for your great work16:28
Keybuk0.10 is going to be simultaneously simpler and more complex ;-)16:29
Keybukie. by simplifying the core design, you can do more complex things with it16:29
Steveethat sounds great16:30
Steveewill there also be that kind of support for "profiles" and / or "cron" replacement ( time based events ) implemented ?16:30
Keybukyes16:31
Steveeyes to both ?16:31
Keybukyup16:31
Steveegreat, thanks a lot16:31
Keybuksimpler > the pre-start/post-start etc. stuff is going away16:31
Keybukso there's just one process per job16:31
Keybukbecause the same functionality is there (in 0.6!) by just doing:16:32
Keybuk  start on starting JOB16:32
Keybuk  script16:32
Steveeatm it's very difficult to provide some kind of rescue mode, or to manage the services on startup16:32
Keybuk    ..16:32
Keybuk  end script16:32
Keybuk(equivalent to a pre-start script)16:32
Keybukso I'm building that on 0.1016:32
Keybukand resulting functionality lets you do things like profiles easily16:33
Steveewill there be any funktion to execute a command on startup a service and on stopping of that service ?16:35
Steveefor example: atm we have the network.conf and we execute as pre-start script "network start" and on post-stop script "network stop"16:35
Steveebut as you told, there could be done more complex things, so may we don't need something like that any more ...16:37
Keybukyes16:39
Keybukyou could define network16:39
Keybukthen16:39
Keybukarg, keep hitting return16:40
Keybukyes, you could define network then use it's "starting" and "stopped" events to put your scripts/exec lines16:40
Keybukyou can do all this in one file in 0.1016:40
Keybuk(at least that's the plan)16:40
Steveewow, great plan / job16:41
Steveei can't await the next release :D16:41
ionNice16:41
KeybukI'm still making sure this all works at the moment ;p16:42
ion:-)16:42
ionI take it there will be both blocking and non-blocking post-start relation?16:43
ionblocking for something that polls until the service is ready and non-blocking for something that should be started after the service and all its blocking post-start scripts are done16:44
ionfor instance16:44
Keybukyeah, that's the bit I'm still working on getting right16:44
Keybukblocking is surprising when you don't want it16:48
Keybukbut not blocking is surprising when you *do* want it16:48
Keybukand it's getting the balance right16:48
ionyeah16:48
Keybukmy current experiment is that "on" blocks and "from"/"while" doesn't16:48
Keybukbut I'm not sure that's right either16:48

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!