| ion | Man, this email is getting longer than i thought. | 00:40 |
|---|---|---|
| twb | As at lucid, "kill -SEGV 1" results in a kernel oops. Is that the intended behaviour? | 06:35 |
| ion | What else should happen? | 06:36 |
| twb | Well, the kernel appears to discard "kill -15 1" and "kill -9 1" | 06:36 |
| twb | I'm OK with it doing that, I was just wondering if it wasn't *supposed* to | 06:37 |
| ralfgro | SpamapS: I need more a condition like start on starting tty1 and (ifstarting kdm or gdm) | 07:17 |
| ralfgro | I would like some feedback on my first shot http://pastebin.com/XxTYPrzS | 07:31 |
| ralfgro | the goal is to start a script before a user can login | 07:32 |
| twb | There's something that deletes /etc/nologin | 07:33 |
| ralfgro | this script has some user interaction, basicially it asks the user if updates can be installed. | 07:33 |
| twb | Work out what it is, hook into that | 07:33 |
| ralfgro | the user should not even has a login screen | 07:33 |
| twb | Then you're probably screwed | 07:33 |
| twb | But you might be able to do whatever fsck does | 07:33 |
| ralfgro | neither console nor gdm/kdm | 07:34 |
| ralfgro | the 3 upstart script I have now do work, at lease in my limited setup | 07:34 |
| ralfgro | but the way I do it seems not be very clean | 07:35 |
| SpamapS | twb: right, I think SIGSEGV is important to deliver to even pid 1 so it can be debugged.. despite that it may oops the kernel. | 08:48 |
| SpamapS | twb: that is a kernel issue... sysvinit also oopses on SIGSEGV | 08:48 |
| twb | Okey dokey | 08:48 |
| twb | Actually I am... abusing it to tell LXC that my container's setup script failed :-) | 08:49 |
| twb | I happened to run the script by accident on a non-container, and it oopsed | 08:49 |
| SpamapS | setup script? you're not using cloud-init ? | 08:53 |
| twb | I'm using LXC | 08:53 |
| SpamapS | you should. It has a local metadata tuype. :) | 08:53 |
| SpamapS | which would have the nice side effect that your LXC container could be booted identical to a cloud node. :) | 08:54 |
| twb | LXC containers don't "boot"; they don't have a kernel. | 08:54 |
| twb | That's rather the point | 08:54 |
| SpamapS | booting starts processes under the container | 09:08 |
| twb | Bah | 09:10 |
| SpamapS | just depends on what you need | 09:10 |
| twb | I should also clarify that by "setup script" I mean the script that takes my company-specific LXC template and applies container-specific changes, like installing apache2 and git cloning a web app into /var/www/. | 09:11 |
| === ion_ is now known as ion | ||
| === apachelogger_ is now known as apachelogger | ||
| foresto | Hi there. I'm hoping someone here can help me. I have upstart launching my daemon using a pretty simple config file with "start on filesystem", but my daemon complains that the nfs-mounted directory it needs does not exist when upstart launches it. I wish I knew how to make it depend on the filesystem that contains the necessary directory. Any suggestions? | 19:45 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!