exitdescription | ok! can anyone tell me again why there's upstart and not the usual startup scripts? | 07:52 |
---|---|---|
exitdescription | hi there, can anyone tell me where the upstart scripts are located? | 08:03 |
exitdescription | okie found it | 08:06 |
exitdescription | thanks! | 08:06 |
ralfgro_ | howdy! | 14:32 |
ralfgro_ | I'm trying to write an upstart script that needs user interaction | 14:32 |
ralfgro_ | it should run after network is up and before any user can log in | 14:33 |
ralfgro_ | so the parallel upstart boot process should wait until this script has finished | 14:34 |
ralfgro_ | I haven't found much on this topic | 14:34 |
mbiebl | ralfgro_: requiring user input during boot is a bad idea | 14:42 |
ion | Interface with plymouth. Its job is to multiplex user interaction for processes. | 14:44 |
ralfgro_ | well, the problem is that not all system are running a DM | 14:45 |
ralfgro_ | we want to update clients during the boot process | 14:45 |
ralfgro_ | but the user has the option to say no | 14:46 |
ralfgro_ | that'S the policy, user can discard updates one time | 14:46 |
ralfgro_ | and we don't have simple laptops or desktops | 14:47 |
ralfgro_ | some systems can't be updated after the have bootet | 14:47 |
ralfgro_ | so we need a way to do it right after network is up | 14:47 |
ralfgro_ | it's getting worse, it has to work with old sysvinit (lenny), insserv (squeeze) and upstart (ubuntu) | 14:48 |
mbiebl | sysvinit/insserv is basically the same | 14:49 |
ralfgro_ | but at the moment I'm trying to find out how it could be done with upstart | 14:49 |
ralfgro_ | I haven't found a way to write an upstart script that simply halts the boot process. I know that's not the goal of upstart, but is this possible? | 14:51 |
mbiebl | ralfgro_: you can ship a sysvinit/LSB script | 14:51 |
mbiebl | upstart still supports sysv init scripts | 14:51 |
ralfgro_ | do you have a link for this? | 14:53 |
ralfgro_ | http://upstart.ubuntu.com/ | 14:53 |
ralfgro_ | was not very helpful | 14:53 |
mbiebl | ralfgro_: if you want to run it before a user can login you have a problem | 14:54 |
mbiebl | because the gdm init script triggers on the filesystem and dbus events | 14:54 |
ralfgro_ | I've seen that in the dgm/kdm scripts | 14:55 |
mbiebl | you could block dbus from starting | 14:56 |
mbiebl | by writing a job that has "start on starting dbus" | 14:56 |
ralfgro_ | can I emit something from an sysv scrip? | 14:56 |
ion | Or just block gdm by ‘start on starting gdm’ | 14:56 |
mbiebl | sure | 14:56 |
ralfgro_ | then I could add that to the start on ... part | 14:56 |
mbiebl | ion: he would also need to block the gettys | 14:57 |
mbiebl | and gdm is started rather late during boot | 14:57 |
mbiebl | so a lot of services are already running | 14:57 |
ralfgro_ | I begin to think it was not the best idea to do updates while booting | 14:57 |
mbiebl | ion: does gdm respect /etc/nologin? | 14:58 |
ion | dunno | 14:58 |
plautrba | in fedora i would use something like 'start on starting start-ttys, console owner, task' | 14:58 |
mbiebl | (looking at /etc/pam.d/nologin, it should) | 14:58 |
mbiebl | /etc/pam.d/gdm, I meand | 14:59 |
ralfgro_ | kdm: auth required pam_nologin.so | 14:59 |
mbiebl | ralfgro_: creating a file /etc/nologin should then prevent users from being able to login | 14:59 |
ralfgro_ | hm, but if gdm is started they don't see the dialog | 15:00 |
ralfgro_ | Update now: y/n.... | 15:00 |
ralfgro_ | this is getting complicated | 15:01 |
mbiebl | ralfgro_: why don't you use the regular update mechanisms | 15:01 |
ralfgro_ | we use puppet for regular desktops/laptops | 15:01 |
mbiebl | so? | 15:01 |
ralfgro_ | but we have a lot of systems that can only be updataes before they are running | 15:01 |
mbiebl | example? | 15:01 |
ralfgro_ | so the people that use these systems needed a special solution | 15:02 |
ralfgro_ | some systems are build in cars | 15:02 |
ralfgro_ | most of the time they are only reachable by net during boot up | 15:02 |
ralfgro_ | others start calculations and can't be updated during normal run time | 15:03 |
ralfgro_ | so need a solution that works for them | 15:03 |
mbiebl | so if the system is up and running it is cut of the network? | 15:04 |
ralfgro_ | probably | 15:04 |
mbiebl | I'd change my deployment to use something like fai or d-i preseeding to have up-to-date systems after installation | 15:05 |
ralfgro_ | the biggest chance is during boot tiem | 15:05 |
ralfgro_ | we install our systems with dai | 15:05 |
ralfgro_ | fai | 15:05 |
ralfgro_ | but we can't use softupdates to keep them up to date | 15:05 |
ralfgro_ | it is very complicated here, everything is a moving target | 15:06 |
ralfgro_ | after install, the systems are modified from the users (yes they have roo). and yes, it's a nightmare | 15:06 |
ralfgro_ | anyway, we have to find a way to install security updates | 15:07 |
mbiebl | I'd try the nologin + start on starting gdm approach then | 15:07 |
ralfgro_ | desktops/laptops get them triggered by puppert at 4am | 15:07 |
mbiebl | (although it is not pretty) | 15:07 |
ralfgro_ | ok, I'll have a look into that | 15:08 |
mbiebl | you probably need "console owner" | 15:08 |
mbiebl | if you use want to prompt via a shell script | 15:08 |
ion | Better to use plymouth. | 15:09 |
mbiebl | if they use it, yeah | 15:09 |
ralfgro_ | on ubuntu yes, ii plymouth 0.8.2-2ubuntu2.2 | 15:10 |
ralfgro_ | but not on debian AFAIR | 15:10 |
ralfgro_ | the goal should be somethin that runs on both | 15:11 |
mbiebl | it is not mandatory, but installable | 15:11 |
ralfgro_ | maybe it would be better and easier to start something right after the user logged in with *GM or console | 15:12 |
ralfgro_ | **DM | 15:12 |
mbiebl | ralfgro_: it should be easier on Debian as the boot process is still linear | 15:12 |
mbiebl | sure, you can use a desktop autostart file for that | 15:13 |
mbiebl | ralfgro_: just run apt-get update during boot | 15:15 |
mbiebl | update-manager should automatically prompt you if there are new updates availabe | 15:15 |
ralfgro_ | we don't want to use update-manager | 15:16 |
ralfgro_ | we need some more logging | 15:16 |
mbiebl | like /var/log/dpkg.log ? | 15:16 |
ralfgro_ | so we will start our own script, if the user decides to install updates on his own, that's fine. but we can't relay on that | 15:17 |
ralfgro_ | so the user can start update-manager, but we trigger our own script | 15:17 |
mbiebl | I still don't understand why you want to cook up your own solution, but ok | 15:18 |
ralfgro_ | as I wrote above, the user have root right. which is needed because the clients are development machines etc. | 15:18 |
ralfgro_ | so sometime they change the sources.list | 15:18 |
mbiebl | i don't see a problem here | 15:19 |
ralfgro_ | because they don't have access to our network when the are abroud | 15:19 |
mbiebl | prompting during boot doesn't solve that problem | 15:20 |
ralfgro_ | but these sources are not reachable from the company network when they are back | 15:20 |
ralfgro_ | in the script we use differnt sources lists witch are fixed | 15:21 |
ralfgro_ | but trust me, it's compicated and we just try to do what is possible in this environment | 15:21 |
ralfgro_ | thanks for the pointers so far | 15:22 |
JanC | why not have a sources list with a hostname that always resolves to a working IP depending on whether you are inside or outside the company LAN? | 16:02 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!