| === Md_ is now known as Md | ||
| === PaulePan1er is now known as PaulePanter | ||
| lnykryn | not quite sure if this is correct place to ask, but did you use some newer autoconf in last version of libnih? https://bugzilla.redhat.com/show_bug.cgi?id=925821 | 11:55 |
|---|---|---|
| xnox | lnykryn: we recommend running autoreconf -f -i before starting the configure/build. | 12:05 |
| lnykryn | xnox: yes, I will do that, but my primary question was more about "official upstream solution" | 12:07 |
| xnox | lnykryn: I'm uploaded libnih in Debian and Ubuntu and I am upstream developer of upstart. We did switch to running autoreconf as part of daily builds & packages in Ubuntu. | 12:08 |
| xnox | lnykryn: I think we will continue releasing tarballs with generated autofoo, but there is no pressing need to release libnih at the moment. | 12:09 |
| xnox | s/uploaded/uploader/ | 12:09 |
| lnykryn | xnox: ok | 12:10 |
| tobin | can someone here shed some light on an upstart issue I'm having. Im trying to run a script at startup (after networking is up), which seems to work, and then when the host gets a shutdown signal, run the script again, with a different arg | 15:29 |
| tobin | https://gist.github.com/joshuatobin/5eb21e379f377825044b | 15:29 |
| tobin | The shutdown part is where i keep running into issues. I've tried various methods with stop on starting network-services etc | 15:29 |
| tobin | but the scrip never seems to get called before networking is shutdown | 15:29 |
| marrusl | tobin, you also tried "start on deconfiguring-networking" iirc. I thought that would work since that's supposed to be a hook. | 15:38 |
| tobin | yeah. I've tried a bunch of different incantations | 15:38 |
| jodh | tobin: you mentioned on #ubuntu-server that your problem was due to 'virtual networking devices'? | 15:38 |
| tobin | jodh: no I just thought maybe that could be a contributing factor | 15:39 |
| jodh | tobin: so, does the log file I mentioned exist? | 15:39 |
| tobin | it does not | 15:39 |
| tobin | Looking to see how to change kernel options now with ec2 | 15:39 |
| tobin | so i can get better logging | 15:39 |
| jodh | tobin: there is a chance the system is shutting down too fast for the job to run to completion. You could convince yourself of that by adding a sleep before the call to "do_stop" in /etc/init.d/sendsigs. | 15:42 |
| tobin | yeah. it seems to shutdown crazy fast (and reboot) | 15:43 |
| tobin | thanks jodh i'll give that a shot as well | 15:43 |
| marrusl | so I guess tobin's job might be blocking deconfiguring-networking, but deconfiguring-networking being blocked won't block shutdown from continuing? | 15:46 |
| SpamapS | marrusl: it will | 16:01 |
| marrusl | yeah, i couldn't figure out why that wasn't working for him then. | 16:01 |
| SpamapS | tobin: your post-stop must finish for deconfiguring-networking to return to the shutdown scripts | 16:02 |
| tobin | The post-stop should be running before network is removed right, in theory? | 16:02 |
| tobin | in my script that is | 16:02 |
| SpamapS | tobin: I'm not sure if the kill timeout applies to post-stops .. You might want to split this into two tasks | 16:02 |
| SpamapS | tobin: since these are definitely tasks, not a single job | 16:03 |
| tobin | okay. Logging isn't working either in post-stop to >> /tmp/upstart.log | 16:03 |
| tobin | works fine in pre-start | 16:03 |
| SpamapS | tobin: well just let 'console log' do its thing, it will log to /var/log/upstart/$jobname.log | 16:03 |
| tobin | SpamapS: okay. gothca | 16:03 |
| tobin | SpamapS: so if i create a task to stop. would the logic be 'start on started deconfiguring-network'? | 16:04 |
| SpamapS | tobin: start on deconfinguring-networking | 16:04 |
| SpamapS | tobin: and the 'task' keyword is critical | 16:04 |
| tobin | okie | 16:05 |
| tobin | thanks i'll try that | 16:05 |
| SpamapS | tobin: also your script needs to be careful about not taking forever. I think, actually, because of how 'deconfiguring-networking' stops the whole process.. anything that starts on it *MUST* exit in a reasonable time or shutdown will never happen. | 16:05 |
| tobin | okay good to know. | 16:06 |
| * SpamapS would love to have an extra keyword for tasks, like 'task seconds 5' to just kill off a task after 5 seconds | 16:06 | |
| tobin | SpamapS: any other suggestions other than deconfiguring-network? I still can't seem to get the shutdown bits in the right order. https://gist.github.com/joshuatobin/3205a6b89d6ea01298c0 | 16:46 |
| SpamapS | tobin: this is very wrong | 17:57 |
| SpamapS | start on (local-filesystems and net-device-up IFACE!=lo) | 17:57 |
| SpamapS | tobin: I know you see it in other jobs, but its a bug almost always | 17:57 |
| tobin | okay. that one works. its the start on deconfiguring | 17:57 |
| tobin | so either way running exec on shutdown keeps screwing me | 17:58 |
| SpamapS | tobin: what do you want to do before the network is down? | 17:58 |
| tobin | run my script, which deletes an A record from route53 | 17:58 |
| tobin | ./ec2_set_hostname.py stop does this | 17:58 |
| SpamapS | tobin: this should work.. http://paste.ubuntu.com/5655866/ | 18:00 |
| SpamapS | tobin: also for the 'start' use 'start on runlevel [2345]' (though with ec2 instances I'm pretty sure the one you have is fine) | 18:03 |
| SpamapS | tobin: just a thought, btw, I would do the manipulation of EC2 from outside the instance. Your scheme is really prone to errors. CloudFormation might be interesting to you. | 18:04 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!