| === apachelogger is now known as releaselogger | ||
| === releaselogger is now known as apachelogger | ||
| zaufi | hi all. anyone alive& | 22:19 |
|---|---|---|
| zaufi | ?? | 22:19 |
| mbiebl | only zombies around here | 22:23 |
| zaufi | :) | 22:23 |
| zaufi | I've got a problem w/ writing upstart script for my daemon... :( anyone can help? | 22:24 |
| mbiebl | not without a crystal ball | 22:28 |
| mbiebl | unless you tell what the problem is | 22:28 |
| zaufi | `service stop mydaemon` won't send a SIGTERM to my daemon... it hust hangs on poll() some socket (dbus I guess)... | 22:29 |
| zaufi | s,hust,just, | 22:29 |
| zaufi | I use 'expect daemon' in the script (which is means that daemon will fork() tiwce, as my daemon do) | 22:30 |
| mbiebl | you mean "stop foo", not "service stop foo"? | 22:30 |
| zaufi | yep | 22:31 |
| zaufi | ... I'm playing with ubuntu 10.10 where `service` calls `start`/`stop` for upstarted programs | 22:31 |
| mbiebl | can you run your daemon without forking? | 22:32 |
| ion | Don’t use the expect stanza unless you’re *absolute* sure of the forking behavior of your main process. The current version of Upstart gets confused if the behavior differs from what the expect stanza claims. | 22:32 |
| ion | absolutely | 22:32 |
| zaufi | I'm prettu sure my daemon fork() exactly twice :) | 22:32 |
| ion | Please pastebin the job definition. | 22:33 |
| mbiebl | does status foo list the correct pid? | 22:33 |
| ion | The next major release of Upstart will have a considerably better method of following forks. | 22:34 |
| zaufi | mbiebl: oops... I'm forget to check... let me reboot to ubuntu and check :) (I'm in gentoo usually) | 22:34 |
| zaufi | I'm just trying to get familiar w/ ubuntu ) | 22:34 |
| zaufi | mbiebl: PIDs are differ... | 22:39 |
| mbiebl | that's your problem then | 22:40 |
| ion | Please pastebin the job definition. | 22:40 |
| zaufi | mbiebl: http://pastebin.com/HRfuCdWr | 22:40 |
| mbiebl | as said, can you run your daemon without forking? | 22:40 |
| zaufi | http://pastebin.com/VHW8UzcV | 22:41 |
| zaufi | Ok I'll try... but how to 'stop` it now :(( | 22:41 |
| mbiebl | kill 2060 | 22:42 |
| zaufi | so I have to comment 'expect' stanza and remove -d option (to allow fg execution) ... am I correct?? | 22:43 |
| mbiebl | yeah, try that | 22:43 |
| zaufi | huh... `start` hangs | 22:44 |
| zaufi | cut it poll() smth | 22:44 |
| zaufi | s,cut,cuz. | 22:44 |
| zaufi | if I Ctrl-C it next start it says that indexer already running | 22:45 |
| zaufi | I guess I have to reboot... to cleanup dbus state | 22:45 |
| zaufi | ?? | 22:45 |
| zaufi | mbiebl: w/o '-d' and expect start/stop works fine | 22:50 |
| mbiebl | problem solved | 22:53 |
| zaufi | not exactly... ( | 22:54 |
| zaufi | my program must be running w/ -d | 22:54 |
| zaufi | in fg mode it makes spam to console only... in -d (daemon) to syslog | 22:54 |
| zaufi | so being running w/o -d from upstart I can't c any log messages | 22:55 |
| mbiebl | zaufi: you can try export fork | 22:57 |
| mbiebl | instead of daemon | 22:57 |
| mbiebl | argh, expect fork | 22:58 |
| zaufi | it fork() twice! | 22:58 |
| zaufi | as most real daemons do | 22:58 |
| ion | Or strace -f it to see exactly what it *really* does. :-) | 22:58 |
| zaufi | `expect fork` won't stop it, as expected | 23:00 |
| zaufi | http://pastebin.com/ZgHubxM8 | 23:04 |
| zaufi | here is a strace result... I used 'expect daemon' | 23:05 |
| zaufi | and it still pilling... | 23:06 |
| ion | I mean strace -o strace.out -f /…/indexer -d | 23:06 |
| zaufi | ok | 23:07 |
| zaufi | just a sec | 23:07 |
| zaufi | this is first child: http://pastebin.com/Vs3EKTbz | 23:13 |
| zaufi | this is second generation child: http://pastebin.com/cKSusw2W | 23:13 |
| zaufi | I used -ff for strace cuz daemon loads JVM (which makes a lot of spam) | 23:14 |
| ion | Use -f | 23:14 |
| zaufi | ok | 23:15 |
| zaufi | 270K ... same file sharing required | 23:16 |
| zaufi | just a sec I'll find smth | 23:16 |
| ion | Perhaps just egrep -A5 -B5 'clone|exec|exit' or something like that. | 23:18 |
| zaufi | http://up.k10x.net/webgkjidzbojj/indexer.log | 23:18 |
| ion | Java seems to do ten forks or so before your code daemonizes. | 23:21 |
| ion | Upstart happily follows the first one of them and gets confused. | 23:21 |
| zaufi | it forks threads | 23:21 |
| zaufi | not processes | 23:21 |
| zaufi | I think I can delay JVM loading... and do it after daemonize... | 23:22 |
| zaufi | I need some tome to fix a code | 23:22 |
| ion | You could separate the parameter to daemonize from the parameter to log to syslog. | 23:23 |
| Voting | I want to start my application level server (a tomcat instance) once everything else is running... how do I run something at the end of the existing startup sequence ? | 23:39 |
| Voting | I want to start my application-level server (a tomcat instance) once everything else is running... how do I run something at the end of the existing startup sequence ? | 23:58 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!