| karshan | hey | 00:06 |
|---|---|---|
| karshan | i'm looking for a way to get an upstart script running on shutdown, the only way I can seem to do it is with a "start on runlevel 0" | 00:06 |
| karshan | is there no analog event to startup in upstart ? | 00:06 |
| JanC | karshan: I thin kthat currently there is none | 00:09 |
| karshan | oh ok.. thanks. you know about any plans of removing the sysv legacy stuff like runlevels or is upstart gonna keep them | 00:09 |
| JanC | karshan: technically, the runlevel stuff isn't part of upstart | 00:10 |
| karshan | well i mean the runlevel events and the stuff in util/sysv.c etc. | 00:10 |
| JanC | some people (especially in the embedded space) don't use any of that | 00:11 |
| karshan | yeah I guess. I was just confused since all /sbin/shutdown does is set the runlevel to 0 and let out an event, it expects some sctript to run /sbin/halt on runlevel 0. | 00:12 |
| JanC | 'shutdown' is part of the sysv compatibility tools IIRC ;) | 00:12 |
| karshan | ahh. yes ofcourse | 00:14 |
| karshan | I still think there should be a shutdown event (not runlevel 0) though, so when something like reboot is executed all jobs are stopped etc. | 00:15 |
| JanC | karshan: in general, upstart will just stop all jobs if you do that | 00:16 |
| karshan | JanC: well what about tasks that need to be carried out like unmounting the filesystems | 00:16 |
| JanC | karshan: you can use whatever event you want for that? | 00:19 |
| JanC | the thing with upstart is that it's very generic | 00:21 |
| karshan | but how do i get the event generated at shutdown | 00:21 |
| JanC | which is an advantage and a disadvantage | 00:21 |
| karshan | i.e. without putting an "inttctl emit shutdown" in the stop script of some random job | 00:22 |
| JanC | karshan: that's exactly what you want to do ;) | 00:23 |
| JanC | actually, put it into the utility/script that starts shutdown | 00:23 |
| karshan | yeah that makes more sense | 00:24 |
| karshan | thats what I was saying though | 00:24 |
| JanC | and whatever name you give it is your choice | 00:24 |
| karshan | shutdown, or since thats sysv legacy, reboot should have something like that | 00:24 |
| karshan | i mean in reboot.c, there should be some code to emit a shutdown event, since it seems pretty useful, most systems would have scripts running on shutdown no ? | 00:26 |
| JanC | it might be useful to add an example reboot which does that | 00:27 |
| karshan | yeah | 00:27 |
| JanC | for those who want to set up a system without syv compatibility | 00:28 |
| JanC | sysv | 00:28 |
| JanC | (I'm not sure if one exists) | 00:28 |
| karshan | mm hmm.. you think its worth posting the code to the mailing list, see if it gets included maybe | 00:29 |
| JanC | Keybuk: ^^^ | 00:29 |
| JanC | karshan: in any case, aking on the ML doesn't hurt ;) | 00:29 |
| JanC | asking | 00:29 |
| JanC | currently most people seem to thing upstart is only upstart + sysv-compatibility as used in Ubuntu | 00:31 |
| karshan | yeah, so i've seen, it took me a while to find any scripts that had nothing to do with sysv compatibility | 00:31 |
| JanC | I expect Google's Chrome OS uses it differently | 00:32 |
| JanC | and the embedded users, but those might not always be public | 00:32 |
| JanC | or more difficult to find | 00:32 |
| karshan | yeah.. i'm building an lfs system, and i didnt want any sysv init stuff, so hopefully i'll have a nice set of upstart scripts legacy free by the end of it | 00:34 |
| JanC | karshan: in that case you can offer them to the project | 00:34 |
| karshan | yup :) | 00:34 |
| JanC | even if only to put them into an "examples" or "contrib" directory | 00:34 |
| Keybuk | hey was at the gym | 00:59 |
| JanC | Keybuk: karshan thinks it would be nice to include examples for a non-sysv-compatible system | 01:01 |
| Keybuk | JanC, karshan: right, the reason there's no non-SyS V shutdown event is because the people using Upstart without the SysV stuff are largely in the embedded space | 01:01 |
| Keybuk | and for them "poweroff -f" is pretty close | 01:01 |
| Keybuk | so it's not been a feature request someone's wanted enough to provide a patch | 01:02 |
| dcorbin_wk | How should (can I?) configure certain upstart services to be controllable by non-root users? | 12:12 |
| JanC | dcorbin_wk: configure sudo to allow them to execute the necessary commands? | 15:39 |
| dcorbin_wk | JanC: the commands are "start" and "stop", right? can you make sudo only allow certain arguments? | 16:01 |
| JanC | yes, see sudoers(5) | 16:01 |
| JanC | there are all sorts of examples in it | 16:03 |
| dcorbin_wk | OK. Thanks. | 16:16 |
| marrusl | Anyone know... can you change the shell that upstart uses for scripts without changing the /bin/sh -> dash link? | 18:24 |
| SpamapS | marrusl: init/paths.h:#define SHELL "/bin/sh" | 18:50 |
| SpamapS | marrusl: tho Keybuk might know another way. ;) | 18:50 |
| marrusl | SpamapS, well THAT I'm not going to support. :-) | 18:50 |
| Keybuk | hmm? | 18:52 |
| SpamapS | Keybuk: " < marrusl> Anyone know... can you change the shell that upstart uses for scripts without changing the | 18:54 |
| SpamapS | /bin/sh -> dash link?" | 18:54 |
| Keybuk | right | 18:54 |
| Keybuk | or | 18:54 |
| Keybuk | ./configure CPPFLAGS=-DSHELL=/bin/crush | 18:54 |
| marrusl | heh. well thanks SpamapS, Keybuk. good to know, it's a weird request. I don't think they'll care enough to want to roll their own. :) | 18:56 |
| Keybuk | it's the kind of change that's best done at compile time | 18:59 |
| marrusl | Keybuk, and they are free to change the link if they'd really like to. Not that i'm suggesting they do. | 19:01 |
| Keybuk | yeah, exactly | 19:06 |
| Keybuk | and really, Why? | 19:06 |
| marrusl | Keybuk, I'm curious, I didn't hear why yet. I know they were working on a network startup issue (LP bug: 777193). they probably added some debug statements that had a bashism in them. probably wasn't anything more than that. | 19:40 |
| radix | I'm trying to get ubuntu working in a vserver... I've got it configured to some extent, where upstart is services properly when it comes up, but after that any start/restart/stop commands I run just block while polling on the connection to init | 20:28 |
| radix | is there something I should try to help figure out what's going wrong? | 20:28 |
| JanC | radix: there are some instructions about upstart + vserver on the vserver wiki IIRC | 21:52 |
| radix | JanC: yeah, and I followed them | 21:52 |
| radix | I'm not convinced that whoever wrote them got it fully working, or maybe just some software has changed since then | 21:52 |
| radix | they got me to the point that services were starting at boot, but I've always had this blocking problem | 21:52 |
| JanC | I have seen people in here who got it working with those instructions | 21:52 |
| * JanC has no experience with it though | 21:53 | |
| Keybuk | if those commands block, you haven't got upstart working in a vserver | 21:53 |
| Keybuk | and are probably using some other daemon | 21:53 |
| reiddraper | if I have a process that I manage with upstart, and it forks on `reload`, will upstart keep track of it? | 21:53 |
| Keybuk | reiddraper: no | 21:53 |
| radix | it's definitely upstart, I've straced both init and the commands | 21:53 |
| radix | and seen them connecting on @/com/ubuntu/upstart | 21:53 |
| reiddraper | Keybuk: is there a workaround? | 21:53 |
| radix | could a misbehaving service specification possibly make one of those commands block? | 21:54 |
| JanC | radix: in some cases | 21:56 |
| radix | 'strace -f restart cron': http://pastebin.ubuntu.com/603486/ 'strace -f -p 1': http://pastebin.ubuntu.com/603487/ | 23:52 |
| radix | could this perhaps offer some clue about my problem? | 23:52 |
| radix | it looks like the processes are at the very least communicating | 23:54 |
| radix | and netstat does tell me that process 1 is listening on @/com/ubuntu/upstart, so I assume that means it is actually upstart :) | 23:55 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!