ikey | Hi - is there an equivalent to systemd-analyze with upstart to determine system readiness+boot time? | 15:35 |
---|---|---|
xnox | ikey: no, but you can install and use bootchart to give you graphs and logs. | 15:36 |
xnox | ikey: (the release series before it became systemd only) | 15:36 |
ikey | xnox, unfortunately the test I'm using wont work with that :) | 15:36 |
ikey | xnox, it boots VMs using qemu in parallel and determines actual length of boot | 15:36 |
ikey | Using a serial device (virio) communicating with a host unix socket | 15:37 |
ikey | Unfortunately the LTS doesnt have systemd support so it wouldnt be a fair test to inject it from a PPA | 15:37 |
xnox | ikey: well that's easy, just feed your console logs to a timestamping logger, e.g. syslog and check what / when happens. | 15:37 |
xnox | ikey: boot with --verbose kernel arg to see more details from upstart as to when/what starts. | 15:38 |
xnox | ikey: or just install bootchart from the archive into that vm. | 15:38 |
ikey | xnox, righto. So the way it works now is systemd tells me the system is up, by systemd-analyze returning a 0 retcode | 15:38 |
ikey | I can't rely on scp back to host because of handshake lag and unreliable networking in Linux in general :) | 15:38 |
ikey | This is starting usually around 20 VMs together and needs a script to run in the guest to echo to the serial0 | 15:39 |
ikey | Is there a way that upstart can run a script *last* when upstart considers the system booted? | 15:39 |
xnox | ikey: yes. | 15:39 |
ikey | oo | 15:39 |
ikey | That'll at least make the test fair :) | 15:39 |
xnox | ikey: that would be equivalent of cloud-init final. Let me give you the sample script. | 15:40 |
ikey | Thanks :) | 15:40 |
xnox | ikey: so, if you are using cloud-init then at the end of boot cloud-init final is executed with "fully booted / ready" messages on the console. | 15:41 |
xnox | http://bazaar.launchpad.net/~upstart-devel/upstart/upstart-jobs/view/head:/utopic/etc/init/cloud-final.conf | 15:42 |
ikey | Thank you very much :) | 15:42 |
xnox | ikey: otherwise you want to add a job "start on stopped rc RUNLEVEL=[2345]" which is after all runlevel 2 tasks have been completed. | 15:42 |
ikey | ah | 15:42 |
xnox | ikey: and system is fully booted, users can login, and things like display-managers / auto-logins are happening... | 15:42 |
ikey | Right | 15:42 |
xnox | and in that job you can send whatever notification you need/want whereever. | 15:43 |
ikey | The way it currently works is incredibly trivial | 15:43 |
ikey | echo "VMREPORTID:%d|||`systemd-analyze` > /dev/virtio-ports/serial0 | 15:43 |
ikey | where %d is taken from a boot parameter in /proc/cmdline | 15:43 |
ikey | Hopefully the virtio-port is up when my script runs ^^ | 15:44 |
* ikey had that with systemd.. | 15:44 | |
xnox | there is no useful "systemd-analyze" but you can do: start on stopped rc RUNLEVEL=[2345] \n exec echo "VMREPORTID:%d|||`uptime` > /dev/virtio-ports/serial0 | 15:45 |
ikey | That sounds fair enough to me | 15:46 |
ikey | Thanks for your help :D | 15:46 |
xnox | ikey: but that's not the same. Also one can add more time entries / points elsewhere. E.g. by simply adding /etc/init/job.override with post-start "uptime after foo" | 15:46 |
ikey | Going to need to give this a good look into :) | 15:46 |
xnox | ikey: no problem. I should look into systemd-analyze and check if that can be made to run against upstart =) | 15:46 |
ikey | Would be cool tbh | 15:46 |
xnox | ikey: i wonder if you are ikey i'm thinking you are. | 15:47 |
ikey | I think it uses an internal signal though | 15:47 |
ikey | heh | 15:47 |
ikey | ..facepalm | 15:47 |
ikey | just whois'd you xD | 15:47 |
xnox | ikey: so you are =) | 15:47 |
ikey | am indeed, howdy :) | 15:47 |
xnox | ikey: well only one week to go. Starting on the 11th =) | 15:48 |
ikey | oh nice :) | 15:48 |
ikey | I go on vacation around the 15th im going to need to check to be sure but perhaps a meet up in london wouldnt hurt | 15:48 |
xnox | ikey: right. And i'll be off to debconf the week after 22-1 =/ | 15:50 |
xnox | ikey: meh, i'm sure we'll stay in touch. | 15:50 |
ikey | Oh ok | 15:50 |
ikey | heh | 15:50 |
ikey | well, hope you enjoy debconf :) | 15:51 |
ikey | oh its in portland this year | 15:51 |
xnox | ikey: yeah, trying to blag auke into keeping me there for an extra week to work from hillsborogh =) | 15:54 |
ikey | haha | 15:54 |
xnox | ikey: he says something rather "budget" and "approval" or some such excuses =))))))))) *giggle* | 15:55 |
ikey | ^^ | 15:55 |
ikey | somewhat envious - wouldnt mind being in portland myself :) | 15:55 |
xnox | ikey: yeah. well not sure. maybe i'll end up going for debconf only / conference leave. And then both of us will go in march for the OTC meet up. | 16:00 |
ikey | possible | 16:01 |
ikey | Anywho I can't take myself to a debconf - I betrayed Debian :P | 16:02 |
xnox | ikey: oh, traitors make the best DDs ;-) come to dark side, we have swirls =) | 16:07 |
ikey | lolz | 16:07 |
ikey | right i gotta get going - meeting soon. ttyiab xnox :) | 16:13 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!