| tartar | I wonder, though, if writing to console in upstart scripts in Lucid could block? | 00:46 |
|---|---|---|
| tartar | I hear plymouth owns the console and bug 785242 implies that writing to it may cause a dead lock. | 00:47 |
| SpamapS | tartar: you should only have one 'console owner' active at one time | 00:54 |
| SpamapS | tartar: you can have some other console owner after plymouth exits | 00:55 |
| === simensen_ is now known as simensen | ||
| cwillu_at_work | I want to set up a job that will ignore its start-on stanza based on a test, but can still be started manually via "start foo-job" | 15:08 |
| cwillu_at_work | is there an environment variable or something I can check to distinguish? | 15:08 |
| SpamapS | cwillu_at_work: pre-start is for that | 15:08 |
| SpamapS | cwillu_at_work: in pre-start, you can just run 'stop' and the job's start will be cancelled | 15:08 |
| cwillu_at_work | sure | 15:09 |
| cwillu_at_work | but that would also trigger if I ran "start foo", wouldn't it? | 15:09 |
| SpamapS | cwillu_at_work: the env variable to check is '$UPSTART_EVENTS' | 15:09 |
| cwillu_at_work | ah, thanks | 15:09 |
| SpamapS | cwillu_at_work: it will be empty with a manual start/stop | 15:10 |
| cwillu_at_work | perfect | 15:13 |
| tartar | thanks, SpamapS | 16:54 |
| tartar | I guess this prevents from easily debugging upstart scripts by echo-ing messages | 16:55 |
| === tartar_ is now known as tartar | ||
| tartar | Although the hardest cases that need debugging may lose that tmpfs file on reboot. | 16:59 |
| SpamapS | tartar: you can have console output as many times as you want | 17:02 |
| SpamapS | tartar: only console owner is dangerous | 17:02 |
| SpamapS | tartar: and newer upstarts log everything to /var/log/upstart now | 17:02 |
| tartar | But I thought plymouth could block these writes to the console when it makes itself busy. The console output lockup bug made me think that "owning" the console meant plymouth injecting itself into the data flow between /dev/console and the screen. I guess I mis-interpreted the bug. (I disliked the newer upstart logging into separate files from each conf file. This makes observing the timeline difficult). | 17:08 |
| SpamapS | tartar: yes plymouth can block, but its job is to not block, and multiplex things properly. | 17:08 |
| SpamapS | tartar: cat /var/log/upstart/* | sort works | 17:09 |
| SpamapS | tartar: the date/time stamps are ISO and sort perfectly | 17:09 |
| tartar | By "blocking" I mean stopping the upstart scripts where they try to write to stdout. | 17:09 |
| tartar | Thanks for the sort trick. I did not see timestamps in some Ubuntu 12 setup. | 17:10 |
| SpamapS | tartar: some things don't have timestamps actually | 17:11 |
| SpamapS | tartar: so you're right about that ;) | 17:12 |
| SpamapS | perhaps upstart should have that as an option | 17:12 |
| SpamapS | tartar: I'm on crack. I spot checked 2 things, and they had timestamps, but those two things were both openstack components which log with a nice structured logger | 17:12 |
| SpamapS | tartar: its just raw "whatever the programs printed" | 17:13 |
| SpamapS | slangasek: ^^ possible upstart feature idea.. add timestamps to logs | 17:13 |
| slangasek | SpamapS: yes, I've been thinking we ought to do that. Could you file a bug report so we don't lose it? | 17:25 |
| slangasek | (and/or provide a patch? :) | 17:26 |
| SpamapS | looking at both right now | 17:29 |
| SpamapS | https://bugs.launchpad.net/upstart/+bug/1154207 | 17:30 |
| xnox | if my crazy idea works, it will be wonderful. | 19:21 |
| SpamapS | xnox: in the US we say it like this: "Hey, watch this!" | 19:25 |
| SpamapS | [boom] | 19:25 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!