File size: 3,697 Bytes
4aa5fce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
[02:54] <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"..
[02:55] <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
[02:56] <MarcWeber> Keybuk: Maube automount should umount -l /auto and exit?
[02:56] <MarcWeber> I don't know which party to patch.
[02:58] <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
[02:59] <MarcWeber> What is the default signal being send to the job to make it quit?
[02:59] <Keybuk> TERM usually ;)
[03:04] <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"
[03:05] <MarcWeber> What is OTHERJOB?
[03:05] <MarcWeber> When I say stop that job I want it to stop.
[03:06] <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
[03:07] <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
[03:08] <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..
[03:09] <MarcWeber> if you don't it will exit.
[03:09] <Keybuk> how does it know?
[03:10] <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
[03:11] <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
[03:12] <MarcWeber> So you umount everything but root
[03:12] <Keybuk> yes
[03:27] <MarcWeber> How long will upstart wait until it sends SIGKILL (if there is no on stopping script running?)
[22:40] <ion> keybuk: Gotta beam off some poop. Start running around. http://www.youtube.com/watch?v=l4HMCCspbUE