mdales | is it possible to have upstart start job b if job a failed? | 12:01 |
---|---|---|
Keybuk | yes | 12:10 |
keesj | hi | 12:19 |
Keybuk | mdales: the stopping and stopped events have arguments saying that it failed, what failed and why | 12:21 |
mdales | ah, excellent | 12:21 |
mdales | is that in the version that's in gutsy? | 12:21 |
mdales | I was reading up on jobfailedevent, but I assume that's not there at the moment. | 12:22 |
mdales | thanks for the info | 12:22 |
Keybuk | it's there | 12:23 |
Keybuk | job-failed-event was implemented in 0.3.x | 12:23 |
mdales | ah | 12:23 |
mdales | w00t | 12:23 |
keesj | I have a job that was started and exite sucessfully , it is not a rewawn job , but I would like the job to remain in "start(ed)" status or at least would like to know that is was started | 12:33 |
keesj | (I currently have a while sds ; sleep 10 ; to keep it running | 12:33 |
Keybuk | I'm not sure I follow | 12:53 |
keesj | :P | 12:54 |
Keybuk | you have a job | 12:55 |
Keybuk | and you want that job to always be running? | 12:55 |
keesj | what I really want to know is if is was stated regardless of whether it is still running | 12:56 |
Keybuk | if it was started, the goal will always be start? | 12:56 |
keesj | much like the mount-kernel-filesystem job | 12:57 |
Keybuk | usual way for that kind of thing is to put the work in the pre-start bit of the job | 12:57 |
keesj | Keybuk: what is the "goal" of a job? | 12:57 |
Keybuk | spore scott% sudo status tty1 | 12:58 |
Keybuk | tty1 (start) running, process 4576 | 12:58 |
Keybuk | the goal is start, which means Upstart will attempt to keep the job running | 12:58 |
Keybuk | if the goal is stop, Upstart will attempt to keep the job _from_ running | 12:58 |
Keybuk | goal is changed to start by events or manual command | 12:58 |
Keybuk | goal is changed to stop by events, manual command and the main process exiting (unless respawn and not normal exit) | 12:59 |
keesj | yes but I don't want it to be running, even if I call start mount-kernel-filesystems | 12:59 |
Keybuk | jdong: I was thinking about whether there should be the option of "sticky" jobs | 13:00 |
Keybuk | that stayed running even after the process exited | 13:00 |
Keybuk | but I couldn't decide whether that was any different from what you were doing, and just doing the work in pre-start | 13:00 |
keesj | after that (because the jobs is no respawn) the status is stop | 13:00 |
keesj | I would like it to be something else :p | 13:00 |
Keybuk | respawn would keep it at start | 13:01 |
Keybuk | (though you would still get stopping/stopped/starting/started events each time it exited) | 13:01 |
keesj | but I don't want that. I want the job to finish gracefully and the status to remain on something else then "stopped" | 13:02 |
Keybuk | the best way to do that is to put your "job" in pre-start | 13:02 |
Keybuk | it'll be run when you start the job, and the job will stay in running (since there's no script/exec) until otherwise stopped | 13:02 |
keesj | the that sounds very sexy! | 13:03 |
keesj | let me try that! | 13:03 |
Keybuk | it's an interesting pattern | 13:09 |
keesj | I can not say how happy I am ! it is working! | 13:27 |
keesj | well kinda ... | 13:37 |
keesj | apparently I can not simply start such a service using emit (with no --no-wait) | 13:40 |
keesj | perhaps it is missing some events | 13:40 |
Keybuk | start? | 13:40 |
* Keybuk really must fix that confusion | 13:40 | |
Keybuk | if you have /etc/event.d/foo | 13:41 |
Keybuk | then the right way to start it is just "start foo" | 13:41 |
Keybuk | emit foo won't do anything unless foo has "start on foo" in its job definition | 13:41 |
keesj | yes , I have things like start on start_boot_mode_select | 13:47 |
keesj | also with start foo I have to use --no-wait, , I am using 0.3.8 | 14:02 |
=== kylem_ is now known as kylem | ||
=== keesj_ is now known as keesj |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!