=== erkules_ is now known as erkules | ||
=== soren_ is now known as soren | ||
Md | I am trying to debug an ubuntu 10.04 system which hangs at boot. If look at what upstart is doing in a different console I see that it started a few processes but I am left with only plymouth running | 15:15 |
---|---|---|
Md | what can I do to find out what's wrong? | 15:15 |
jodh | Md: take a look at http://upstart.ubuntu.com/cookbook/#debugging. | 15:25 |
jodh | Md: first, I'd try booting with --debug on the kernel command-line and capturing the log output to see if anything looks suspicious. | 15:25 |
jodh | Md: have you installed any new jobs recently or changed any default jobs maybe? | 15:25 |
Md | jodh: this is a customer server and I have no idea if and how they broke it | 15:25 |
Md | the problem does not appear to be a job hanging, unless it's plymouth itself | 15:26 |
Md | I already checked the log with --verbose | 15:26 |
jodh | Md: do you know if this server has ever booted successfully? | 15:26 |
Md | jodh: yes, it was working until it was rebooted (or crashed, who knows) | 15:27 |
Md | I can see that the only jobs started are plymouth, hwclock, ureadahead and hostname | 15:28 |
Md | the only possibly suspicious message is two "ignored event 40 (1) for process ..." messages | 15:28 |
Md | where the process is plymouth | 15:29 |
jodh | Md: this will list missing standard jobs: http://paste.ubuntu.com/5798845/ | 15:35 |
jodh | Md: this will list modified jobs: http://paste.ubuntu.com/5798853/ | 15:35 |
jodh | Md: look at the newest files in /etc/init/ and try to identify any new jobs which might be causing problems. | 15:36 |
whg | Is there any sort of "best practice" around managing Python processes inside a virtualenv? | 15:37 |
Md | there are no new files | 15:44 |
whg | Can I source scripts in shell script I'm running under upstart? | 15:45 |
Md | no missing files, no modified files | 15:45 |
jodh | whg: I suggest you read the Cookbook: http://upstart.ubuntu.com/cookbook/#sourcing-files | 15:46 |
whg | Dammit | 15:46 |
whg | I swear I searched the cookbook for that | 15:46 |
whg | Thanks for the link. | 15:46 |
jodh | whg: np. | 15:46 |
jodh | Md: any new jobs? If not, has /etc/fstab been modified? are all expected disks mounted? | 15:47 |
Md | no new jobs, only one disk | 15:47 |
jodh | Md: I think we'll need to see the output from --verbose/--debug then. | 15:49 |
Md | jodh: http://www.bofh.it/~md/u/ | 15:57 |
jodh | Md: looks like mountall might be the problem. Look for any odd lines/options in /etc/fstab. | 16:13 |
Md | there is nothing unusual, the file is trivial | 16:15 |
Md | if I run "initctl start upstart" it mounts / rw | 16:15 |
Md | I mean "initctl start udev" | 16:16 |
jodh | Md: well, from the files you've provided, it doesn't appear that mountall gets as far as spawning. | 16:16 |
jodh | Md: mountall is supposed to do that | 16:16 |
=== ion_ is now known as ion | ||
Md | yes, so what should start mountall? | 16:22 |
Md | if I try to start it manually I get a "Job failed to start" error | 16:22 |
Md | got it | 16:24 |
Md | the problema was a syntax error in /etc/default/locale, which was over one year old | 16:24 |
Md | can you make sure that a || true is added to mountall in the next release? :-) | 16:36 |
JanC | Md: you mean when sourcing /etc/default/locale ? | 16:45 |
Md | JanC: yes | 16:46 |
Md | there are still some issues BTW, only some services have been started | 16:46 |
Md | I can start them manually, but I have no clue why upstart is not starting them | 16:47 |
JanC | considering that that file might get changed on many systems, and admins are humans, so make errors, maybe that wouldn't be a bad idea :) | 16:48 |
Md | nevermind, this one looks like a fuckup by a coworker | 16:48 |
slangasek | Md: ah, that bug was just recently reported, https://bugs.launchpad.net/ubuntu/+source/mountall/+bug/1192514 | 16:58 |
slangasek | wonder why everyone has invalid /etc/default/locale all of a sudden :P | 16:58 |
Md | there was no error message either at boot time or when starting mountall manuall BTW | 16:59 |
slangasek | yes; job output handling only became sensible much later (13.04, I think?) | 17:02 |
JanC | maybe upstart needs a library of functions that provides "best practice" methods for common tasks in upstart jobs... | 18:03 |
JanC | (that also means they can be fixed centrally if they aren't really best practice ;) ) | 18:05 |
xnox | JanC: nah, ideally daemons/software is capable of running with just a single exec line. | 18:05 |
xnox | JanC: without need of any scripts at all. | 18:05 |
xnox | JanC: plus we have http://upstart.ubuntu.com/cookbook/ already. | 18:06 |
xnox | JanC: problem with "helper" libraries is that they quickly gain feature creep trying to accommodate every use case. | 18:07 |
JanC | xnox: ideally, all programs would be bugless :p | 18:07 |
xnox | JanC: please no, otherwise I'd be out of work :P | 18:07 |
JanC | you can always write new bug-free software :) | 18:08 |
JanC | and to avoid too much feature creep, you just need a good gatekeeper ;) | 18:13 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!