| candyban | Hi guys. I'm having an issue with ubuntu 10.04 where some scripts (e.g. snmpd, conntrackd) do not start properly because some interfaces are not ready yet (e.g. eth2/eth3). Both scripts have Required-Start: $network, so I wonder if there is a bug in upstart where it does not wait for the dependency to be met before starting the script | 10:46 |
|---|---|---|
| candyban | anyone alive in here? | 11:14 |
| candyban | Is it possible for upstart to run /etc/rcX.d serial rather than parallel ? | 11:15 |
| candyban | can I force scripts to somehow wait for "all" interfaces to be available & configured before continuing? | 11:16 |
| twb | On Ubuntu 10.04, there are upstart services that I wish guarantee are NEVER EVER started, but due to package dependencies, I can't simply remove the package via apt. | 12:08 |
| twb | Since "update-rc.d foo disable" and policy-rc.d are not applicable, my next instinct is to "dpkg-divert --rename /etc/init/avahi-daemon.conf", so that upstart will see (and hopefully ignore) a file "/etc/init/avahi-daemon.conf.distrib". | 12:08 |
| twb | Is there a better way to blacklist services? | 12:09 |
| candyban | twb, I'm here for help myself on 10.04 ... usually when I have a problem like that, I go ugly and just do chmod -x /usr/sbin/avahi-daemon | 13:11 |
| candyban | twb, or I suppose you can edit /etc/init/avahi-daemon.conf | 13:12 |
| twb | Permissions on non-conffiles aren't preserved | 13:12 |
| twb | But yeah, you could dpkg-divert /usr/sbin/avahi-daemon | 13:13 |
| candyban | twb, just add exit 0 at the top of /etc/init/avahi-daemon.conf (not sure if that works) | 13:13 |
| twb | Unfortunately, upstart jobs are a hodge-podge of DSL and embedded sh | 13:14 |
| twb | So I doubt that would work. | 13:14 |
| twb | In any case, editing the file is basically what I'm trying to avoid. | 13:15 |
| oli1980 | Hi all | 13:16 |
| oli1980 | i have a strange issue with upstart :/ when trying to stop ssh in Ubuntu 10.06 the stop command does not return | 13:17 |
| soren | twb: Just remove the job? | 13:23 |
| twb | soren: you mean rm -f it? | 13:23 |
| twb | Hmm, I suppose that WOULD work, because it's a conffile | 13:24 |
| soren | twb: Yes, that's what I mean. | 13:24 |
| soren | twb: It's absolutely supported. And absolutely offtopic for this channel, but there you go :) | 13:25 |
| twb | Sorry! | 13:25 |
| twb | I guess because upstart jobs (i.e. conf files) are distro-specific, and thus off-topic? | 13:25 |
| soren | twb: Conffile handling is not an upstart issue. | 13:46 |
| twb | Righto | 13:46 |
| oli1980 | when i try to stop ssh with "stop ssh" i simply get in the debug of init init: job_process_handler: Ignored event 20 (5) for process PID | 13:49 |
| candyban | soren, is there a way to start the network and make all the rest wait until that is completed? | 14:42 |
| soren | candyban: Generally, no. | 14:44 |
| soren | candyban: Mostly because there's no way to know when you're going to be done configuring the network. | 14:45 |
| soren | candyban: I mean.. It's very difficult to define when configuration of the network has been completed. | 14:45 |
| candyban | soren, I'm having trouble with snmpd (which I bind on eth3) and conntrackd (which syncs on eth2) | 14:46 |
| soren | candyban: then start them when they come up. | 14:46 |
| candyban | soren, they are both started before the network is properly initialized | 14:46 |
| soren | candyban: "start on net-device-up IFACE=eth2" | 14:47 |
| soren | Adjust as appropriate. | 14:47 |
| soren | This is assuming an Ubuntu system, by the way. | 14:47 |
| soren | Other distros may or may not have a net-device-up event. | 14:47 |
| candyban | soren, 10.04 (but snmpd and conntrackd are "regular" init scripts in /etc/init.d ... there is no equivalent in /etc/init ... do I need to recreate the scripts? | 14:48 |
| soren | candyban: You can also add a post-up line to your network configuration in /etc/network/interfaces that just calls the relevant init scripts. This is completely off-topic here, though. If you want to discuss this, please take it somewhere like #ubuntu-server | 14:51 |
| candyban | soren, I never had these kinds of problems with regular SysV boot ... so I spent most of today learning about upstart as I want to understand what and where it goes wrong and how I can fix it the right way. (e.g. creating a new upstart script for snmpd/conntrackd) | 14:54 |
| soren | No, with sysv boot you had different problems. | 14:54 |
| candyban | soren, true, but at least it was predictable. Now booting of certain services fails/succeeds at random | 14:55 |
| candyban | soren, anyways, back to topic. Should I create a new /etc/init script (or can I modify the existing SysV-style script) | 14:56 |
| soren | I'm not sure what you want me to say. Even driven boot is racy if not configured correctly. All you need to do is configure it correctly. | 14:56 |
| soren | As I said: This is completely off-topic here, though. If you want to discuss this, please take it somewhere like #ubuntu-server | 14:56 |
| candyban | soren, ok, thanks. | 14:57 |
| bencer | hi all | 18:01 |
| bencer | i'm having some issues with upstart starting squid 2.x on lucid | 18:01 |
| bencer | syslog says: | 18:01 |
| bencer | Jun 25 16:58:50 eboxdev1 init: squid main process (27716) killed by TERM signal | 18:01 |
| bencer | Jun 25 16:58:50 eboxdev1 init: squid main process (28441) terminated with status 1 | 18:01 |
| bencer | but squid after his default shutdown_lifetime says: | 18:01 |
| bencer | 2010/06/25 16:59:21| Squid Cache (Version 2.7.STABLE7): Exiting normally. | 18:02 |
| bencer | it's like upstart was killing squid for not being stopped inmediately | 18:02 |
| bencer | and look at the times, it matches the default 30sec squid shutdown_lifetime | 18:03 |
| bencer | and the funny thing is that if i comment the pre-script on the upstart script i can't reproduce the issue | 18:05 |
| bencer | i think i've found that the expect fork option doesn't work very well, just not using that and running squid with -N (do not background) works as expected | 18:28 |
| bencer | going to submit a patch on squid package... | 18:28 |
| === notting_ is now known as notting | ||
| === notting_ is now known as notting | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!