| [diablo] | moin moin | 02:02 |
|---|---|---|
| [diablo] | anyone about? | 02:02 |
| [diablo] | morning all | 08:15 |
| [diablo] | anyone about please? | 08:15 |
| JanC | [diablo]: if you have a question, ask it, and then stay around--if somebody knows the answer, they will answer when they are online & have time | 08:19 |
| [diablo] | hi JanC | 08:19 |
| [diablo] | sorry, reason I asked it was dead here when I last tried | 08:19 |
| [diablo] | ok here goes | 08:19 |
| [diablo] | basically I have a problem with ubuntu 10.04 not starting qemu-kvm and libvirt-bin on boot | 08:20 |
| [diablo] | it only started after I changed my NIC to bridge mode | 08:20 |
| [diablo] | anyway.... I started looking at the rc.N ... all is set correctly.. | 08:20 |
| [diablo] | then I checked the upstart jobs, yep.. all there | 08:20 |
| [diablo] | but... I want to start debugging it (upstart) | 08:21 |
| [diablo] | I look at man init | 08:21 |
| [diablo] | and I see it says in the files section at the bottom | 08:21 |
| [diablo] | FILES | 08:21 |
| [diablo] | /etc/init.conf | 08:21 |
| [diablo] | /etc/init/*.conf | 08:21 |
| [diablo] | but there is no /etc/init.conf file | 08:21 |
| [diablo] | and I want to work out the boot order of the services that upstart does on boot | 08:22 |
| JanC | if there is no init.conf it's not used ;) | 08:22 |
| [diablo] | ohhhh | 08:22 |
| [diablo] | k | 08:22 |
| [diablo] | so, the order of the services starting??? | 08:22 |
| JanC | so, only look in /etc/init/*.conf | 08:22 |
| [diablo] | nod | 08:23 |
| JanC | well, that depends on the scripts in that directory | 08:23 |
| [diablo] | ok is there a primary script, first point of execution? | 08:23 |
| JanC | you might want to read init(5) (run "man 5 init") | 08:24 |
| [diablo] | ok | 08:25 |
| JanC | basically, it starts services based on events | 08:25 |
| [diablo] | ok | 08:26 |
| ion | I seem to remember a known bug regarding bridge initialization in Ubuntu. You’ll probably find it on Launchpad. | 08:26 |
| [diablo] | JanC, the bridge is working actually | 08:27 |
| [diablo] | JanC, the problem is that qemu-kvm and libvirt-bin must now be started manually once the machine has booted | 08:27 |
| JanC | lookign at their init scripts, it seems like they should start on startup | 08:29 |
| JanC | well, they should start on runlevel 2, 3, 4 & 5 | 08:30 |
| [diablo] | actually, due to faffing around with update-rc.d ... I have sysv links | 08:31 |
| [diablo] | I will remove them | 08:31 |
| [diablo] | gonna reboot the box now | 08:33 |
| [diablo] | and see... | 08:33 |
| JanC | there should be a symlink /etc/init.d/libvirt-bin -> /lib/init/upstart-job normally | 08:34 |
| [diablo] | nod | 08:34 |
| JanC | (which implements sysv-compatibility) | 08:34 |
| JanC | and LSB-compatibility, I suppose | 08:35 |
| [diablo] | nope, LSB nope | 08:35 |
| [diablo] | ok, just rebooted... | 08:35 |
| [diablo] | nada de nada ... still neither service is running | 08:35 |
| JanC | does your hardware support kvm? | 08:36 |
| [diablo] | this is weired | 08:36 |
| [diablo] | yep sure does.. if I manually start the services, all is good | 08:36 |
| [diablo] | and previously it all worked, it was only when I changed to bridge that these two services stopped booting | 08:37 |
| JanC | ion: maybe try what ion said: he thinks there is a known bug about that ;) | 08:38 |
| [diablo] | in the bridge-utils package ion ? | 08:38 |
| [diablo] | sorry, was thinking it was you (JanC) who said it | 08:38 |
| JanC | might be in libvirt or whatever | 08:39 |
| [diablo] | ok, but that should not effect qemu-kvm | 08:39 |
| [diablo] | as that just does a modprobe | 08:39 |
| JanC | and it doesn't modprobe? | 08:40 |
| JanC | modules aren't loaded? | 08:40 |
| [diablo] | nope | 08:40 |
| [diablo] | but if I manually run it now | 08:40 |
| [diablo] | diablo@beast:~$ lsmod |grep kvm | 08:41 |
| [diablo] | diablo@beast:~$ sudo service qemu-kvm start | 08:41 |
| [diablo] | [sudo] password for diablo: | 08:41 |
| [diablo] | qemu-kvm start/running | 08:41 |
| [diablo] | diablo@beast:~$ lsmod |grep kvm | 08:41 |
| [diablo] | kvm_intel 46296 0 | 08:41 |
| [diablo] | kvm 286392 1 kvm_intel | 08:41 |
| JanC | /etc/default/qemu-kvm has "KSM_ENABLED=1" ? | 08:42 |
| JanC | although, that shouldn't change things on manual start | 08:42 |
| [diablo] | yep | 08:43 |
| [diablo] | enabled | 08:43 |
| [diablo] | it´s gotta be something with the bridge | 08:43 |
| ion | The bug might be against the upstart package, or even the udev package. | 08:43 |
| ion | or ifupdown | 08:43 |
| [diablo] | hi ion | 08:43 |
| [diablo] | well, it only started once I changed to bridge | 08:44 |
| [diablo] | will just check launchpad about what your saying ion | 08:45 |
| JanC | [diablo]: maybe also ask in #ubuntu-virt and/or #ubuntu-server | 08:48 |
| JanC | https://bugs.launchpad.net/ubuntu/+source/qemu-kvm/+bug/580454 might be the bug ion was thinking about | 08:48 |
| [diablo] | thanks JanC | 08:54 |
| [diablo] | looks about right | 08:54 |
| [diablo] | well | 08:55 |
| [diablo] | mmm sort of | 08:55 |
| [diablo] | but I will just have to fire the two services up manually until a resolution is found | 08:55 |
| [diablo] | many thanks for you help JanC and ion | 08:55 |
| [diablo] | all the best | 08:55 |
| hallyn | on this vostro 1220, in order to get it to respect 'auto eth1' (eth1 being broadcom sta wireless) in /etc/networks/interfaces, i had to add 'pre-up exec ifdown -a' to /etc/init/networking.conf | 14:53 |
| Psi-Jack | Oh cool. There's an entire channel just for upstart eh? | 15:40 |
| Psi-Jack | Anyway. I'm trying to learn about upstart, and the documentation for it, is minimal at best so far that I can see. What I'm trying to work with now is respawn and limit. If I understand it right, basically by using respawn, a watchdog tracks the process and respawns it should it tie by abnormal means without using upstart process control, correct? | 15:43 |
| mbiebl | Psi-Jack: upstart (/sbin/init) tracks the process | 15:47 |
| Psi-Jack | Right. :) | 15:47 |
| Psi-Jack | So if I flagged respawn on, for example, my keepalived upstart-job, and I killall -9 keepalived, upstart should respawn it basically correct? | 15:48 |
| ion | Yep | 15:49 |
| Psi-Jack | How's the limit definition work? | 15:49 |
| ion | See init(5) | 15:50 |
| Psi-Jack | Well, that doesn't show much at all. | 15:51 |
| Psi-Jack | Ahh wait. ;) | 15:51 |
| Psi-Jack | Now that, makes a lot more sense. ;) | 15:55 |
| Psi-Jack | With upstart supervising a process, say, keepalived, which I setup so that conntrackd and ipvsadm are linked with keepalived's starting and stopping events, if I killed keepalived abnormally, would conntrackd ad ipvsadm follow with it, or would they remain running? | 15:56 |
| ion | I think they’ll be restarted when keepalived is respawned, but better check that for yourself. | 15:57 |
| Psi-Jack | Hmm. I'm testing it now, before adding the respawn part to it, just to see what happends. | 15:58 |
| === notting_ is now known as notting | ||
| Psi-Jack | There is one thing that has me boggled a bit, though. | 16:26 |
| Psi-Jack | When I was making an upstart job for ipvsadm, I had some issues. | 16:26 |
| Psi-Jack | I ended up having to use pre-start script and post-stop script so it wouldn't actually try to track the process itself, but what happends when I start ipvsadm otherwise, it loads it as [ipvsadm_syncmaster] and/or [ipvsadm_syncbackup], instead of as a normal process. | 16:27 |
| Psi-Jack | upstart fails to keep track of that it seems. | 16:27 |
| Psi-Jack | So, is it possible somehow to get an upstart job to basically track a kernel process thread, rather than a normal process daemon or fork? | 16:30 |
| neal | I'm trying to write an upstart job for Maemo 5 (upstart version 0.3.8). My program forks into the background. It seems that this means upstart doesn't know how to find its pid and kill it. (stop ...) Should I use pre-stop and kill it manually? | 16:30 |
| Psi-Jack | neal: expect fork | 16:31 |
| Psi-Jack | If it forks out 1 process, expect fork. If it daemonizes and runs 3 processes, expect daemon | 16:31 |
| neal | I thought that was only 0.5.0+ | 16:31 |
| Psi-Jack | Oh. | 16:31 |
| Psi-Jack | hmm. Maybe. ;) | 16:31 |
| Psi-Jack | Upgrade time. ;0 | 16:31 |
| neal | this is Maemo 5 | 16:32 |
| neal | I can't. | 16:32 |
| Psi-Jack | There is only do, or do not, there is no can't. | 16:32 |
| neal | sure. | 16:32 |
| neal | Any other suggestions? | 16:32 |
| Psi-Jack | I personally only know more of the newer stuff about upstart, so I have nothing. | 16:33 |
| Psi-Jack | without the advancements upstart's gone, I probably wouldn't be using it like I am now. | 16:33 |
| Psi-Jack | Personally, I'm curious how well the cron-like advancements are coming." :) | 16:39 |
| ion | Perhaps with 0.10, which should be released for Ubuntu 10.10 if all goes well. | 16:47 |
| Psi-Jack | That soon eh? | 16:48 |
| Psi-Jack | Well, it'll probably be a while till I switch off 10.04. Probably end up sticking with it till the next LTS. | 16:48 |
| Psi-Jack | Hmm | 20:37 |
| Psi-Jack | Has any progress been made on an apache upstart job script? | 20:38 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!