| twb | As at Ubuntu 10.04 (upstart 0.6.5-8), where do default ulimits come from, when they're not set in the init job? Is it just /etc/security/limits.conf like normal? | 02:11 |
|---|---|---|
| twb | Apparently I have a python daemon that's cracking the shits about only being allowed to open 500 files at once, and I can't see where that's being set on my server. | 02:12 |
| twb | Hm, /proc/N/limits says $user is lying, since that limit is set to 1024/4096 | 02:13 |
| salty-horse | hi. this is a bit off-topic. does start-stop-daemon allow two separate services that use the same <command> in "start-stop-daemon --exec <command>" ? | 10:03 |
| xnox | salty-horse: sure, why wouldn't it? each invocation of start-stop-daemon doesn't know about any other invocations. | 10:07 |
| xnox | salty-horse: make sure you use more parameters to identify which instance is which (e.g. using pidfiles, etc) | 10:08 |
| xnox | jodh: awesome inotify branch ! | 10:08 |
| jodh | xnox: thanks :) | 10:08 |
| salty-horse | xnox, thanks! so should I use --pidfile instead of --exec? what if both commands have the same --exec but different --pidfiles? would that be enough to differentiate them? or does it make no sense to have several "matching" flags? | 10:10 |
| xnox | salty-horse: "--exec foo --pidfile minifoo.pid" and "--exec foo --pidfile bigfoo.pid" are different and thus correct one will be stopped/reloaded/restarted/etc | 10:13 |
| xnox | salty-horse: but e.g. I'd rather recommend you to use upstart & it's support for instances =) | 10:13 |
| salty-horse | xnox, in my case, each so called "instance" has different flags. does this still fit the definition of upstart instances? | 10:16 |
| salty-horse | I'm specifically trying to run both a mongodb "regular server" and a mongodb "config server". someone copied the upstart conf file, and modified it, but now both files have the same --exec, and only one of them has a --pidfile. so they stepped on each other's toes and I could not start both at the same time | 10:17 |
| salty-horse | I sense setting a --pidfile on both is the simplest solution | 10:17 |
| xnox | salty-horse: it should, but it all depends on your usecase. E.g. /etc/init/network-interface.conf defines "instance $INTERFACE" thus one gets a running job per each configured network interface. There are other examples, just grep for "instance" in /etc/init/ | 10:18 |
| xnox | salty-horse: using a human-identifiable "instance $NAME" will get you what you want, e.g. "production", "staging", "testing", "dev" ...... mongodb servers =) | 10:19 |
| xnox | but it kind of assumes that you can structure all your mongodb servers in a homogenic way. E.g. for each of them you start up with --config /etc/mondonbd/$NAME.config | 10:19 |
| xnox | or shell you have shell to select what ever you need based on the instance name =) | 10:20 |
| xnox | (e.g. any options, ports, etc....) | 10:20 |
| salty-horse | xnox, from what I'm reading about isntances, they require an extra parameter on the startup command. I'd rather not deal with variables. I want a simple "service NAME start" | 10:23 |
| xnox | salty-horse: then create different jobs for each one. Or you can create wrapper jobs: e.g. configmongodb.conf will simply have "start mongodb NAME=config", and all the common-logic will be in the mongodb.conf | 10:34 |
| xnox | salty-horse: that way you aid maintainance, by not repeating the same snippet in each job. | 10:35 |
| xnox | salty-horse: or have normal jobs, without instances, but source shell in pre-start script if there is anything significant to share between each mongo jobs. | 10:36 |
| salty-horse | that seems overkill for my current problem, but I appreciate the knowledge for future configurations. if a job only has "start ..." does upstart still knows how to shut it down correctly? | 10:37 |
| xnox | salty-horse: good point. i think one would also need similar stubs for "stop/restart/reload mongo NAME=config" then in pre|post-start|stop =( | 10:39 |
| salty-horse | :) | 10:39 |
| xnox | possibly something like 5-6 lines in total for the whole conf. | 10:39 |
| xnox | cause one would simply test what the goal is in "pre-start" & "pre-stop" & "post-start" | 10:39 |
| xnox | *sigh* | 10:40 |
| salty-horse | thanks a bunch. sorry for depressing you about upstart inflexibility. at least now you can add a new keyword that does that magic | 10:52 |
| xnox | salty-horse: nah, it's ok =) if I complain about upstart, I'm the one who has to implement features in it ;-) | 11:04 |
| leandrosansilva | Hello. Is it possible to create custom events? I created a service called service1 and put the code "emit my_event" and I also created another one called service2 and put "start on my_event". But, when I start service1, service2 doesn't start. Probably I'm doing something (everything) wrong, but what? | 11:28 |
| leandrosansilva | Currently I'm reading the upstart cookbook, but there's much do be seen there | 11:28 |
| xnox | leandrosansilva: instead of emitting & catching events you can simply do: "start service2" in service1.conf, or in service2.conf "start on started service1" | 11:29 |
| xnox | leandrosansilva: can you please pastebit full jobs. Did you declare that service1 emits events using "emits" stanza? http://upstart.ubuntu.com/cookbook/#emits | 11:30 |
| leandrosansilva | The problem this is just a test. I really don't need these services. In fact I need to start a service as soon the network starts (it's a daemon which needs to access some external services on the local network), and I'd like it started on event static-network-up, but it didin't work | 11:31 |
| leandrosansilva | and I don't want to change the networking service | 11:31 |
| leandrosansilva | oks, just a minute | 11:31 |
| xnox | leandrosansilva: please note that once an event is emitted - it's gone, adding new jobs will not "magically notice the event" | 11:34 |
| xnox | you can download upstart-gui to monitor the events emitted on the buses. | 11:35 |
| leandrosansilva | So I think I must solve my problem in another way | 11:35 |
| leandrosansilva | what I need is just this: my daemon must start just after the network is up | 11:35 |
| xnox | leandrosansilva: please paste your jobs, or describe your problem. | 11:35 |
| leandrosansilva | and I wouldn't like to change networking.conf to start my daemon there | 11:36 |
| xnox | leandrosansilva: "start on (local-filesystems and net-device-up IFACE!=lo)" | 11:37 |
| leandrosansilva | my idea is making my daemon (called capturad) depends on networking | 11:37 |
| xnox | that means start onces file-systems got mounted & there is any kind of network connection (non-localhost only) | 11:37 |
| xnox | leandrosansilva: documented at http://upstart.ubuntu.com/cookbook/#normal-start | 11:38 |
| leandrosansilva | oh, it's easier than I thought | 11:38 |
| leandrosansilva | I'll try this | 11:38 |
| leandrosansilva | a moment... | 11:38 |
| xnox | leandrosansilva: well you need to reboot to test / generate those events properly =)))) | 11:39 |
| leandrosansilva | hum.. So it may be the problem | 11:39 |
| leandrosansilva | I haven't restarted | 11:40 |
| leandrosansilva | xnox, wIth a simple daemon, "start on (local-filesystems and net-device-up IFACE!=lo)" seems to work. Let me check with the real daemon. Anyway, thanks a lot for your help! | 11:45 |
| xnox | no problem | 11:45 |
| leandrosansilva | xnox, I've tested with the real daemon and it worked correctly. Thx again | 11:58 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!