| === apachelogger is now known as releaselogger | ||
| robs_ | someone can take a look at this bug I submitted ? https://bugs.launchpad.net/upstart/+bug/703993 | 11:35 |
|---|---|---|
| ronny | hi | 12:57 |
| ronny | im a bit at loss, i cant seem to find docs on how to prevent the automated starting of some services | 12:58 |
| JanC | ronny: what version of upstart are you using? | 13:02 |
| JanC | ronny: in 0.6.7 you can use the "manual" stanza, in older versions you can comment out the "start on" line(s) | 13:03 |
| ronny | atm 0.6.6 | 13:08 |
| dacresni | how can you which running process depend on an upstart job | 15:20 |
| dacresni | like what's using mysql now | 15:20 |
| JanC | considering that those processes might be running on a computer at the other side of the world, how do you think that would work? | 15:25 |
| dacresni | other side of the world? im just talking about programs that only start after mysql is running | 15:28 |
| dacresni | thats the point of upstart right? to only launch programs after the programs they depend on are runnning? | 15:28 |
| dacresni | in the case of mysql if there are are open sockets i guess i should ask the mysql people about that right? | 15:29 |
| dacresni | or am I confusing this with launchd | 15:29 |
| jhunt | dacresni: it's not that easy currently - you need to look at the "start on" and "stop on" entries for all jobs. | 15:35 |
| JanC | upstart is based on events, not on dependencies | 15:36 |
| jhunt | quite. | 15:37 |
| JanC | maybe it's possible to document dependencies in job metadata in the future, just like you can document "emits" ? | 15:38 |
| JanC | dacresni: programs/jobs that need mysql might not always have a connection/socket open permanently... | 15:40 |
| dacresni | ok, thanks | 15:41 |
| dacresni | the events vs dependencies tells me that i did confuse it with launchd | 15:43 |
| dacresni | I forget how systemd does it | 15:43 |
| JanC | you can implement dependencies using upstart's events, but not the other way around | 15:44 |
| dacresni | well, thanks all | 16:02 |
| === ion___ is now known as ion | ||
| === releaselogger is now known as apachelogger | ||
| kreign | hi, I'm trying to understand the mountall.conf file. I need to have it wait for a custom filesystem module, which is getting modprobed via module-init-tools.conf; however, both these conf files appear to have the same prereq of 'start on startup' | 22:16 |
| kreign | is there any way to make one a prereq of the other? | 22:16 |
| kreign | from looking at the files it looks like I can do an "emits fs_module" and then have mountall.conf with "start on (startup and started fs_module)" - is this correct? | 22:18 |
| JanC | "emits" is documentation | 22:18 |
| JanC | so it does nothing on its own | 22:19 |
| kreign | JanC, ok; how can I specify that basic chain of events? | 22:20 |
| kreign | JanC, start on (startup and started init-module-tools) ? | 22:21 |
| kreign | er module-init-tools | 22:21 |
| JanC | sounds like that might work | 22:22 |
| kreign | JanC, or am I going about this the wrong way? | 22:22 |
| kreign | i'm assuming it's not whitespace sensitive... | 22:22 |
| JanC | I guess the "startup" part can be left out too | 22:22 |
| kreign | hmm probably. | 22:23 |
| JanC | and maybe you can make the modprobe for that specific module its own task, so that mountall can start earlier? | 22:24 |
| kreign | JanC, hmm I don't think mountall will be able to start earlier - module-init-tools is the direct prerequisite. | 22:25 |
| JanC | also depends on whether boot time is important ;) | 22:25 |
| kreign | not so much. | 22:25 |
| * kreign misses the old way of doing things, somewhat. | 22:25 | |
| kreign | JanC, I'd think that any filesystem modules should be loaded before filesystems try to mount, anyway. | 22:26 |
| JanC | normally they will be loaded automatically? | 22:27 |
| kreign | JanC, that does not seem to be the case, in this regard. | 22:28 |
| kreign | JanC, the modules aren't strictly "linux compliant" per se | 22:28 |
| kreign | (ok, not going to lie - it's zfs, and I"ve also got to import the zpool in the module-init-tools.conf) | 22:28 |
| JanC | kreign: I think for most filesystems, their mount utility loads the modules, or something like that | 22:30 |
| JanC | I guess 'mount.zfs' (or whatever it's called) doesn't? | 22:31 |
| kreign | JanC, running mountall manually once in does the trick (with 'auto' as an fstab option) | 22:32 |
| kreign | JanC, it just looks like mountall is running prior to zfs module getting loaded (defined in /etc/modules) and not knowing what -t zfs is. :) | 22:33 |
| kreign | so, it dies. | 22:33 |
| JanC | right | 22:33 |
| JanC | like I said, you can use an extra job only to load the ZFS modules and make mountall.conf depend on that task | 22:35 |
| kreign | I don't recall classic init hanging on exception like this... iirc it'd continue despite problems. | 22:35 |
| JanC | classic init did a lot of things serially | 22:36 |
| kreign | JanC, right - but the zfs modules would need to be loaded beforehand, which are defined in /etc/fstab, so it makes sense to have the module-init-tools.conf do that. | 22:36 |
| JanC | if you don't mind waiting until all those modules are loaded | 22:36 |
| JanC | in most cases not an issue, I think | 22:37 |
| kreign | JanC, yeah. it's on server hardware, so... | 22:37 |
| kreign | slow bios | 22:37 |
| kreign | that's like 3-4 times as long as boot even though it's booting from sloooooow flash media | 22:37 |
| JanC | ☺ | 22:38 |
| kreign | well let's hope i didn't biff it | 22:38 |
| JanC | my desktop doesn't boot fast either ;) | 22:38 |
| kreign | i don't want to walk across the building again | 22:38 |
| kreign | 15:38:28 up 123 days, 23:34, 9 users, load average: 0.66, 0.57, 0.49 | 22:38 |
| kreign | ^ not much a problem for me. | 22:39 |
| kreign | tho i should probably do something like that. I've had two new kernels installed since then... | 22:39 |
| JanC | right | 22:39 |
| JanC | I was going to say: I install security updates :P | 22:39 |
| kreign | JanC, oh I do too. | 22:39 |
| kreign | just don't usually have the time for a reboot... will probably do that tonight and watch monday be unproductive fixing things. | 22:40 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!