radix | for what it's worth, 'stop cron' does actually kill the cron process (as does 'restart cron', though that one doesn't bring it back up) | 00:13 |
---|---|---|
radix | so it's communicating enough that it can do its job, but then not cleaning up properly or continuing to further steps, I guess | 00:13 |
radix | for what it's worth, --verbose doesn't seem to write anything when running stop / restart at runtime, though I do see all the console output at first run | 00:35 |
radix | er, at boot | 00:35 |
=== JanC_ is now known as JanC | ||
radix | did anyone get around to looking at those strace outputs I pastebined yesterday? | 18:39 |
steffen_b | hi there | 18:58 |
steffen_b | searching for some helping hand of shutdown | 18:58 |
steffen_b | with natty | 18:59 |
steffen_b | + debugging | 18:59 |
steffen_b | problem is that a handfull of services is not shutdown | 19:00 |
JanC | define "not shutdown" | 19:04 |
steffen_b | its stopped on runlevel | 19:06 |
steffen_b | 016 | 19:06 |
steffen_b | and i can see until very late in shutdown process normal messages | 19:06 |
steffen_b | of the daemon | 19:06 |
steffen_b | it doesnt do any shutdown messages | 19:07 |
steffen_b | stopping plugins | 19:07 |
steffen_b | or anything | 19:07 |
steffen_b | and late in the boot | 19:07 |
steffen_b | i get stopped with status 1 | 19:07 |
steffen_b | i.e. | 19:08 |
steffen_b | openbox main process (1273) terminated with status 1 | 19:08 |
JanC | I doubt openbox is handled by upstart? | 19:09 |
Keybuk | with that message, I'd say it is | 19:10 |
steffen_b | in my install it is | 19:10 |
steffen_b | i written the job ;) | 19:10 |
JanC | if you are doing such a customized setup, why use runlevels? | 19:11 |
steffen_b | nice try ;) | 19:11 |
steffen_b | stop on runlevel [016] | 19:11 |
steffen_b | is openbox also others which look the same | 19:12 |
steffen_b | the system is ubuntu natty , customized | 19:12 |
JanC | heh | 19:12 |
steffen_b | i dont want to customize everything only the part which we care | 19:13 |
steffen_b | thats why we base it on ubuntu | 19:13 |
steffen_b | i put a test job in to see if the runlevel come | 19:13 |
steffen_b | and that one is executed fine | 19:14 |
JanC | is X also started by upstart? | 19:14 |
steffen_b | the openbox job starts X | 19:14 |
steffen_b | xinit openbox basically | 19:14 |
JanC | so it's probably X that exits with status 1 ? | 19:15 |
steffen_b | i dont debug that one job | 19:16 |
steffen_b | its maybe 10 | 19:16 |
steffen_b | all show the same behaviour | 19:16 |
steffen_b | all working fine on lucid | 19:16 |
JanC | did you check the PID is actually correct? | 19:19 |
steffen_b | believe me the jobs are working | 19:20 |
steffen_b | May 5 19:12:47 vdr init: plymouth-stop pre-start process (1269) terminated with status 1 | 19:20 |
steffen_b | May 5 19:12:47 vdr init: plymouth-splash main process (1328) terminated with status 1 | 19:20 |
steffen_b | May 5 19:39:45 vdr init: vdr-frontend main process (1437) terminated with status 143 | 19:20 |
steffen_b | May 5 19:39:45 vdr init: udisks-automounter main process (1594) terminated with status 143 | 19:20 |
steffen_b | May 5 19:39:45 vdr init: plymouth-upstart-bridge main process (2186) terminated with status 1 | 19:20 |
steffen_b | May 5 19:39:45 vdr init: irexec main process (2245) terminated with status 1 | 19:20 |
steffen_b | this terminated with status 1 doesnt look normal | 19:20 |
steffen_b | and is also not happening each boot | 19:21 |
steffen_b | so something strange is happening in the shutdown | 19:21 |
JanC | doesn't seem to happen on a "normal" natty system either? | 19:22 |
steffen_b | looks like something in the shutdown is killing the jobs while they are active | 19:23 |
steffen_b | May 5 19:39:45 vdr init: irexec main process (2245) terminated with status 1 | 19:23 |
steffen_b | May 5 19:39:45 vdr init: irexec main process (2246) terminated with status 1 | 19:23 |
steffen_b | May 5 19:39:45 vdr init: irexec main process (2247) terminated with status 1 | 19:23 |
steffen_b | if you look at this | 19:23 |
steffen_b | i don't have a "normal" natty | 19:24 |
steffen_b | but i can check that | 19:25 |
steffen_b | i would rather debug whats happening | 19:25 |
steffen_b | then trying empiric if it's my fault or ubuntus ;) | 19:25 |
JanC | wait, I'm seeing things like "Apr 28 13:47:16 saeko init: plymouth-stop pre-start process (3747) terminated with status 1" too here | 19:31 |
steffen_b | or maybe coming from the other side | 19:31 |
steffen_b | if upstart is using dbus for handling the jobs, how can upstart shutdown dbus ? | 19:32 |
JanC | upstart only uses dbus for communication with dbus clients | 19:32 |
JanC | and it doesn't need the dbus daemon | 19:32 |
JanC | for communication with upstart clients I mean | 19:33 |
steffen_b | k , so initctl uses | 19:33 |
JanC | things like initctl | 19:33 |
steffen_b | something else | 19:33 |
steffen_b | not dbus | 19:33 |
JanC | dbus is a wire protocol, you can use dbus without a dbus daemon ;) | 19:34 |
steffen_b | well on lucid i tried to send signals in init.d scripts and they got plain ignored back then | 19:35 |
JanC | using initctl? | 19:35 |
steffen_b | now it looks like runlevel [016] gets ignored | 19:35 |
steffen_b | yes | 19:35 |
steffen_b | which one gets ignored seems to be random/depending on time | 19:36 |
JanC | that's weird | 19:36 |
steffen_b | my requirement is to cleanly shutdown the daemon | 19:36 |
steffen_b | nothing should kill it before it is finished | 19:37 |
JanC | you have a pre-stop script for that? | 19:39 |
steffen_b | no | 19:40 |
steffen_b | only post start | 19:40 |
JanC | default way to stop a job is to send SIGTERM, then after some time-out send SIGKILL, so if shutdown of your service takes longer, you'll need a pre-stop to handle this gracefully I think? | 19:42 |
steffen_b | https://bugs.yavdr.com/projects/yavdr/repository/entry/trunk/base/yavdr-base/etc/init/vdr.conf | 19:42 |
steffen_b | but pre-stop would wait until this is finished | 19:43 |
steffen_b | before actually stopping it | 19:43 |
steffen_b | pre-stop -> kill with SIGTERM vdr doesn't sound right to me | 19:45 |
steffen_b | also it wouldnt help if something would kill it | 19:46 |
steffen_b | from init.d | 19:46 |
steffen_b | i can not believe that natty shutdown is so fucked up or handcrafted for the 5 known use cases | 19:48 |
steffen_b | sry | 19:52 |
JanC | I don't know what exactly goes wrong with your setup, but when using pre-stop for a job you can tell it to shut down and wait until it's ready, after that upstart will see no SIGTERM/SIGKILL is needed for that job anymore | 20:05 |
JanC | but maybe you want to know what goes wrong first... | 20:06 |
steffen_b | there are 2 issues | 20:08 |
steffen_b | first is that runlevel is not coming or sendsigs is coming to fast | 20:08 |
steffen_b | second is waiting for the daemon to finish | 20:09 |
steffen_b | the pre-stop might be nice workaround | 20:09 |
steffen_b | but worth nothing if rc and those sendsigs is killing the jobs | 20:10 |
akio | Hello, I have finally figured out how to handle a broken serial port and got console redirection working but I don't know where to put the fix. | 20:10 |
steffen_b | on the other hand that should ignore everything which is handled by upstart | 20:11 |
akio | I have a ttyS0.conf in /etc/init that starts it for me, but before it starts it needs to be manually configured with setserial. | 20:11 |
akio | I was wondering where to put the setserial command. | 20:11 |
steffen_b | create a setserial.conf | 20:11 |
steffen_b | start on starting ttys0 | 20:12 |
steffen_b | ttyS0 | 20:12 |
akio | Oh cool thanks. | 20:12 |
steffen_b | https://bugs.yavdr.com/projects/yavdr/repository/entry/trunk/base/yavdr-base/etc/init/setserial-minimal.conf | 20:13 |
akio | Man upstart really has made GNU\Linux better in a big way for me. | 20:13 |
akio | Too complicated for me to follow. | 20:14 |
steffen_b | its cool, but causing me headaches sometimes | 20:14 |
akio | Makes boot super fast. | 20:14 |
steffen_b | if its just for you , just but the commands you need in the script section ;) | 20:15 |
steffen_b | start on starting ttyS0 | 20:15 |
steffen_b | will make it wait until that is finished | 20:15 |
steffen_b | you might need to put task keyword also in | 20:16 |
SpamapS | yes, otherwise it won't block properly | 20:16 |
SpamapS | task makes it block until the job has completed the whole lifecycle | 20:17 |
akio | where does task go? | 20:20 |
SpamapS | on its own line | 20:21 |
SpamapS | anywhere tho | 20:21 |
akio | http://pastebin.com/AHPTn0GW | 20:22 |
steffen_b | yes should work | 20:23 |
akio | rebeaut! | 20:23 |
akio | steffen_b: Thank you. That worked like a charm. | 21:39 |
akio | Because charms work well. | 21:39 |
steffen_b | cool :) | 21:39 |
akio | Had an issue last night and couldn't console because previous admin couldn't figure it out. | 21:39 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!