| MarcWeber | How to handle this automount case? If /auto is active and you stop autofs then autofs just prints "Can't shutdown, /auto still active".. | 02:54 |
|---|---|---|
| MarcWeber | However when you release /auto (by cding away) it still doesn't stop. Either upstart should resend some signals or automount should recall that it should shutdown when /auto can be released.. | 02:55 |
| Keybuk | there probably is no event for a filesystem being in use or not | 02:55 |
| MarcWeber | Keybuk: Maube automount should umount -l /auto and exit? | 02:56 |
| MarcWeber | I don't know which party to patch. | 02:56 |
| Keybuk | filesystems are hard | 02:58 |
| Keybuk | they're pretty integral to the "upness" of a system | 02:58 |
| Keybuk | bringing them up and tearing them down has proven quite the tough cookie | 02:58 |
| MarcWeber | What is the default signal being send to the job to make it quit? | 02:59 |
| Keybuk | TERM usually ;) | 02:59 |
| MarcWeber | Maybe I should start this on stopping : "set -e; while :; do pkill -TERM automount; sleep 1; done | 03:04 |
| Keybuk | or have automount "stop on starting OTHERJOB" | 03:04 |
| MarcWeber | What is OTHERJOB? | 03:05 |
| MarcWeber | When I say stop that job I want it to stop. | 03:05 |
| Keybuk | what you were thinking of putting that code into | 03:06 |
| Keybuk | ie. | 03:06 |
| Keybuk | if you have | 03:06 |
| MarcWeber | The problem is that automount get's the signal but ignores it. | 03:06 |
| Keybuk | /etc/init/umount.conf | 03:06 |
| Keybuk | exec umount -a | 03:06 |
| Keybuk | and you need automount stopped first | 03:06 |
| Keybuk | then | 03:06 |
| MarcWeber | The script keeps resending the TERM signal. Upstart sends it only once | 03:06 |
| Keybuk | /etc/init/automount.conf | 03:06 |
| Keybuk | stop on starting umount | 03:06 |
| Keybuk | sure | 03:06 |
| Keybuk | Upstart sends it once | 03:06 |
| Keybuk | waits for automount to get its act in gear | 03:06 |
| Keybuk | if it doesn't follows up with SIGKILL | 03:07 |
| MarcWeber | IF SIGKILL is sent /auto keeps mounted. I guess this is a autostart bug. | 03:07 |
| MarcWeber | It should umount -l then.. | 03:07 |
| Keybuk | yes | 03:07 |
| Keybuk | automount shouldn't ignore SIGTERM | 03:07 |
| Keybuk | how do you tell automount to stop normally? | 03:08 |
| MarcWeber | It doesn't. But it neither unmounts its filesystems. | 03:08 |
| MarcWeber | You send TERM ? | 03:08 |
| Keybuk | you said it ignored TERM :p | 03:08 |
| MarcWeber | IT does if you cd into the /auto directory.. | 03:08 |
| MarcWeber | if you don't it will exit. | 03:09 |
| Keybuk | how does it know? | 03:09 |
| MarcWeber | Don't ask me. Probably it tries umount /auto and notices that that command fails | 03:10 |
| Keybuk | oh | 03:10 |
| Keybuk | right | 03:10 |
| Keybuk | so don't do that then ;) | 03:10 |
| MarcWeber | That's not an option. | 03:10 |
| MarcWeber | I'm a human. I am allowed to make mistakes causing trouble.. | 03:10 |
| MarcWeber | :) | 03:10 |
| Keybuk | usually you kill all processes before unmounting anyway | 03:10 |
| Keybuk | ie. killall5 -TERM; killall5- KILL; umount -a | 03:10 |
| MarcWeber | Keybuk: It's another issue: I'm using NixOS. It restarts the job whenever the configuration changes. So maybe I should write an exception for that job as well .. | 03:11 |
| Keybuk | possibly | 03:11 |
| MarcWeber | hehe. How do you run halt then? | 03:11 |
| Keybuk | after the filesystems are unmounted | 03:11 |
| Keybuk | (you remount root read/only rather than unmounting) | 03:11 |
| MarcWeber | Then the command may be gone.. | 03:11 |
| Keybuk | halt is on the root | 03:11 |
| MarcWeber | So you umount everything but root | 03:12 |
| Keybuk | yes | 03:12 |
| MarcWeber | How long will upstart wait until it sends SIGKILL (if there is no on stopping script running?) | 03:27 |
| ion | keybuk: Gotta beam off some poop. Start running around. http://www.youtube.com/watch?v=l4HMCCspbUE | 22:40 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!