sadmac | Keybuk: just sent an email to our dbus maintainer to get those patches in rawhide | 00:45 |
---|---|---|
Keybuk | who's that? | 00:47 |
sadmac | David Zeuthen | 00:47 |
Keybuk | lol | 00:58 |
Keybuk | I'll nag him tomorrow ;) | 00:58 |
sadmac | very nice | 00:59 |
sadmac | He's in westford too :) | 00:59 |
sadmac | I'll have to bring one of those paper fans. | 01:00 |
=== Amaranth_ is now known as Amaranth | ||
=== Keybuk_ is now known as Keybuk | ||
=== Keybuk_ is now known as Keybuk | ||
sadmac2 | Keybuk: good morning | 15:01 |
Keybuk | afternoon | 15:19 |
sadmac2 | well, you're closer to GMT, so we'll use your clock | 15:20 |
=== sadmac2_ is now known as sadmac2 | ||
sadmac2 | restarting dbus on a running system = epic fail | 16:22 |
Keybuk | yes | 16:23 |
sadmac2 | Keybuk: that reminds me, if I SIGTERM or otherwise force upstart to re-exec, will it transfer job state to the new instance? | 16:23 |
ion_ | s/(?:restarting | on a running system)//g | 16:24 |
sadmac2 | ion_: ORLY? | 16:24 |
ion_ | YA | 16:24 |
sadmac2 | Keybuk: the test case make targets don't seem to run the dbus code generator as a dependency | 16:39 |
suihkulokki | sadmac2: dbus should do that.. dammit if it is possible to transfer the irc server connections when upgrading irssi, bloody well something as critical as dbus should do that as well | 16:41 |
sadmac2 | Keybuk: and yes, your patches seem to fix things. | 16:46 |
sadmac2 | your dbus patches that is | 16:46 |
sadmac2 | Keybuk: wb. Just added another patch to my branch | 18:35 |
sadmac2 | rearranged some conditional logic in job_class_reconsider | 18:35 |
sadmac2 | Keybuk: I see the foo ? TRUE : FALSE; construct in a lot of places. Is Upstart code so sensitive to precise values of TRUE that this is necessary? | 20:55 |
EvanR | hey i just got upstart working on my system, many services are started using the existing sysv scripts... however my scripts print messages during startup like 'samba started.... [OK]', and these get messed up with upstart | 21:27 |
sadmac2 | EvanR: what platform? | 21:27 |
EvanR | linux | 21:27 |
sadmac2 | EvanR: what distro rather | 21:28 |
EvanR | its mostly blfs | 21:28 |
EvanR | with upstart it will be a lot less like blfs | 21:28 |
EvanR | i am about to start moving all the stuff in /etc/rc.d/init.d to event.d scripts | 21:29 |
EvanR | is there a best practice for printing out status messages like that ? | 21:30 |
EvanR | or is it all automatic | 21:30 |
sadmac2 | no practice for printing out messages from job defs | 21:32 |
sadmac2 | dunno why the ones in init.d wouldn't work right. something to do with the FD upstart is passing to the job | 21:32 |
EvanR | does upstart do things in sequential order, or is multi-thread-like stuff going to affect the messages ordering | 21:33 |
sadmac2 | upstart is paralell, but when it is emulating sysv (at least, with the standard jobset which does this) it runs the sysv jobs in serial | 21:34 |
EvanR | i have the example jobs which do rc* compatability | 21:35 |
sadmac2 | yes. those will use your /etc/rc script to run the sysv jobs. | 21:35 |
EvanR | lets say that i convert all my stuff to event.d ... is there a common / best / popular way to implement the runs levels, or is the recommended strategy something different in the upstart worl | 21:37 |
EvanR | like... x-mode text-mode single-user mode events or something | 21:37 |
sadmac2 | upstart does not, by default, use runlevels | 21:39 |
EvanR | is the concept of runlevels a purely sysv init thing or what | 21:39 |
sadmac2 | the replacement for them is kind of up in the air with 0.3.9 | 21:40 |
sadmac2 | 0.5 should see some better mechanisms to deal with different modes. | 21:40 |
EvanR | does it take a specific kernel parameter as input | 21:40 |
EvanR | how would i get to single user mode in emergency | 21:41 |
sadmac2 | Kernel parameters can be gotten at any time from anywhere by reading and parsing a certain proc file. you can do it yourself in a shell script with sed | 21:41 |
EvanR | ok so theres no standard | 21:41 |
sadmac2 | you can have a single job starting on startup that parses kernel parameters from /proc and then triggers another event depending on the result | 21:41 |
sadmac2 | that would do it. | 21:41 |
EvanR | what does ubuntu do at the moment | 21:42 |
sadmac2 | I believe ubuntu is still largely in sysv-compatible mode. | 21:42 |
sadmac2 | though theres a couple community projects that make it use more of upstart's functionality | 21:43 |
EvanR | so back to the issue of bootup status messages, you say its all parallel, whats a good strategy to get those message shown | 21:43 |
sadmac2 | echo from within the scripts has worked for me. I don't know why you would have issues. | 21:44 |
EvanR | ok ill try to get deeper into what is going wrong | 21:44 |
EvanR | trying to write a job for samba. script /usr/sbin/smbd; /usr/sbin/nmbd; end script starts, but immediately stops (leaving the daemons running). what am i doing wrong | 22:42 |
EvanR | /usr/sbin/smbd -D; /usr/sbin/nmbd -F; leaves the job running but sort of hangs the start command, and then stop kills nmbd but not smbd | 22:43 |
EvanR | perhaps i should have smbd and nmbd in two separate jobs | 22:47 |
sadmac2 | EvanR: you need the "respawn" or "service" stanza in your job definition. | 22:48 |
sadmac2 | and you don't want any of your services to daemonize | 22:48 |
sadmac2 | so --nodaemons all around | 22:48 |
EvanR | ok so i should do | 22:49 |
EvanR | /usr/sbin/smbd -F & | 22:49 |
sadmac2 | no & | 22:49 |
sadmac2 | if you use respawn, then start won't hang | 22:49 |
EvanR | i would rather have these two servers controlled by one job | 22:49 |
sadmac2 | No. They should be separate | 22:49 |
EvanR | hmmmm | 22:49 |
EvanR | so i have jobs smbd and nmbd, i would like to be able to say # start samba, stop samba | 22:50 |
sadmac2 | # start smbd | 22:50 |
sadmac2 | (nmbd then figures out it needs to be running and runs) | 22:51 |
EvanR | ok | 22:51 |
EvanR | respawn doesnt sound right | 22:51 |
sadmac2 | you'll just add a "start on started smbd" line to the start of the nmbd job | 22:52 |
sadmac2 | respawn means if it crashes upstart will bring it back. If you don't want that, then use "service" in stead. | 22:52 |
EvanR | oh, service and respawn are mututally exclusive? | 22:53 |
sadmac2 | respawn implies service | 22:53 |
sadmac2 | and actually, your job definition script should end in an exec | 22:54 |
sadmac2 | so "exec /usr/sbin/smbd -F" | 22:54 |
EvanR | i have that | 22:54 |
EvanR | and for nmbd | 22:54 |
sadmac2 | right | 22:54 |
EvanR | well service before that | 22:54 |
EvanR | and "start on started smbd" "exec /usr/sbin/nmbd -F" | 22:55 |
sadmac2 | right | 22:55 |
EvanR | the start on stuff makes things automatic, but what if i want to start jobs conditionally, given what happens in the script of another job | 22:57 |
sadmac2 | like when? | 22:57 |
EvanR | ...........dunno | 22:57 |
EvanR | :P | 22:58 |
EvanR | i cant think of a reason | 22:58 |
sadmac2 | If there's no use case, there's probably no way to do it. If a use case appears, we'll probably think of a way to do it :P | 22:58 |
EvanR | to disable services at bootup, one would comment out the 'start on' stuff... | 22:58 |
bochecha | hi! | 22:58 |
sadmac2 | yes | 22:58 |
EvanR | heh | 22:59 |
bochecha | just came to point a typo on the upstart main site: in the "know users" section, it's written "Fedora Rawhide". It should be "Fedora 9 and above" as Fedora 9 was released :) | 23:00 |
bochecha | not really important, but I saw it while searching for infos about upstart | 23:01 |
sadmac2 | Yes it was :) | 23:01 |
* sadmac2 pats himself on the back | 23:01 | |
bochecha | Anyway, upstart seems like a great project. Keep up the good work! | 23:05 |
* bochecha goes back to read some doc to understand better how it works | 23:06 | |
EvanR | my dbus sysvinit stop script is testing the exit status of the killed daemon to determine whether it should remove a file | 23:24 |
EvanR | is that possible to do from the post-stop script | 23:24 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!