| === harlowja is now known as harlowja_away | ||
| nk121 | Hi #cloud-init. Question about upstart jobs -- is it safe to assume that after all of cloud init stuff is done, it starts the service? | 06:25 |
|---|---|---|
| Odd_Bloke | nk121: "The service"? | 10:53 |
| nk121 | i mean the upstart job | 10:53 |
| Odd_Bloke | nk121: Which job are you referring to? | 10:54 |
| nk121 | an upstart job included in the cloud-init configuration | 10:54 |
| nk121 | does it get started at the end of everything being run? | 10:55 |
| nk121 | it seems to, but i'm not sure when in the lifecycle its started (like is it the last thing that happens when cloud init is run?) | 10:55 |
| nk121 | (so should runcmd and user data scripts assume that the upstart job is not running) | 10:56 |
| nk121 | Odd_Bloke: ^ | 10:58 |
| Odd_Bloke | nk121: I'm not sure how the ordering is determined, I'm afraid. | 10:58 |
| nk121 | Odd_Bloke: That's ok, of course. :) | 10:59 |
| Odd_Bloke | smoser might be able to help out, but he's on EST so won't be around for a few hours. | 10:59 |
| Odd_Bloke | nk121: Examining the code, it looks like there isn't a defined ordering; so things will be executed in the order Python's dict implementation spits them out. | 11:00 |
| Odd_Bloke | But it just takes me missing one line for that to be completely wrong. :) | 11:00 |
| nk121 | Yeah | 11:00 |
| nk121 | hmm, so then its probably not safe to have a dependency on write_files and runcmd (i have a script that is written out in write_files and then executed as a specific user in runcmd) | 11:02 |
| nk121 | so far, it seems to run my runcmd block after the file is created | 11:02 |
| Odd_Bloke | nk121: Aha, OK, I think the ordering is defined by cloud.cfg. | 11:25 |
| nk121 | is that a section? | 11:26 |
| nk121 | or you mean, a file in the source | 11:26 |
| Odd_Bloke | nk121: That'll (probably) live in /etc/cloud; it should be installed by your cloud-init package. | 11:26 |
| nk121 | hmm | 11:28 |
| nk121 | cloud_config_modules: | 11:28 |
| nk121 | # Emit the cloud config ready event | 11:28 |
| nk121 | # this can be used by upstart jobs for 'start on cloud-config'. | 11:28 |
| nk121 | - emit_upstart | 11:28 |
| Odd_Bloke | nk121: So I think the upstart job is written out before any of these things run. | 11:29 |
| nk121 | okay i'll have to continue this in the morning, its way past my bed time (i'm in PST). thanks for your help tonight | 11:30 |
| Odd_Bloke | nk121: Sleep well! :) | 11:31 |
| Odd_Bloke | smoser: Daily reminder for https://code.launchpad.net/~daniel-thewatkins/cloud-init/fix-smartos/+merge/252874 :) | 11:48 |
| smoser | Odd_Bloke, is that python2 safe ? | 12:59 |
| smoser | does ser.readline() return something that can be .decode()ed in python2 ? | 12:59 |
| smoser | nk121, the job will be written pretty early in boot. | 13:00 |
| smoser | so unless you're upstart job is set to run really early, it will be started by upstart. | 13:00 |
| smoser | but cloud-init lets upstart take care of that. | 13:01 |
| smoser | the 'emit upstart' isn't waht does it. | 13:01 |
| smoser | merged | 13:09 |
| Odd_Bloke | smoser: You can decode() anything in Python 2. | 13:11 |
| Odd_Bloke | >>> "".decode().decode().decode() | 13:11 |
| Odd_Bloke | u'' | 13:11 |
| smoser | yeah, just tested that locally. | 13:11 |
| === harlowja_away is now known as harlowja | ||
| suro-patz | logger url | 16:51 |
| suro-patz | logger: url | 16:57 |
| harlowja | suro-patz there is no logger in most public channels; thats typically a y! concept ;) | 17:08 |
| suro-patz | is there a archive for cloud-init, I do not see that in pipermail | 17:09 |
| Odd_Bloke | suro-patz: It is logged to irclogs.ubuntu.com (e.g. http://irclogs.ubuntu.com/2015/03/17/%23cloud-init.html) | 17:10 |
| Odd_Bloke | I _think_ that is updated hourly. | 17:10 |
| suro-patz | cool! thanks Odd_Bloke | 17:10 |
| nk121 | smoser: thanks. Could you clarify/confirm: 1) Is it a safe assumption to be run a file written in a write_files section in a runcmd or user data script? 2) If an upstart job is added via cloud-init, can you assume that its been installed into /etc/init and started before code in runcmd/user_data_script is run? | 19:08 |
| smoser | nk121, http://bazaar.launchpad.net/~cloud-init-dev/cloud-init/trunk/view/head:/config/cloud.cfg | 19:09 |
| smoser | write-files writes the files. | 19:10 |
| smoser | scripts-user in cloud_final_modules is what runs your 'runcmd' provided stuff. | 19:11 |
| smoser | so yes. 1 is safe and guaranteed. | 19:11 |
| nk121 | ok, so for 1) write_files is in init, and runcmd is in config and user-data is in final, so that seems correct | 19:11 |
| smoser | 2 is guaranteed except for if init system is not upstart (which... in vivid it is not and wont be in 16.04) | 19:12 |
| nk121 | oh not, not run-cmd itself? | 19:12 |
| smoser | run-cmd itself actually just writes the command to a script that will later get run. interestingly :) | 19:12 |
| nk121 | sorry i mean run-cmd doesn't run run-cmd | 19:12 |
| nk121 | ah | 19:12 |
| nk121 | ok, so you are saying though i should not assume in the long run that my upstart job will be running when my runcmd/user-data script is run as that will change in a future ubuntu | 19:13 |
| nk121 | within a stage, is that the order htey are run in? | 19:13 |
| nk121 | err, you are implying ubuntu is moving away from upstart? | 19:14 |
| smoser | yes. within a stage they're run in that order. | 19:15 |
| nk121 | oh, systemd replacement | 19:15 |
| smoser | yes, 15.04 will use systemd by default. | 19:15 |
| smoser | and thus 16.04. | 19:15 |
| nk121 | okay, got it. thank you. | 19:16 |
| smoser | i'm not planningon making cloud-init in any version magically convert upstart to systemd jobs | 19:16 |
| smoser | but there will likelyi be in a future cloud-init a way to write systemd jobs like you can write the upstart jobs now. | 19:17 |
| suro-patz | smoser: there? | 22:35 |
| smoser | suro-patz, here for 3 minutes maybe | 23:32 |
| harlowja_ | hmmm, i think he just wanted your feedback on https://code.launchpad.net/~suro-patz/cloud-init/vm-clone-ip-reusage-issue | 23:43 |
| harlowja_ | if u get some time; u seem to have thoughts on that idea; so i'm gonna let u decide if thats ok to let in, lol | 23:44 |
| suro-patz | thanks harlowja_ ! | 23:45 |
| harlowja_ | np | 23:46 |
| harlowja_ | although i think i missed him to, lol | 23:46 |
| harlowja_ | was 10 minutes to late | 23:46 |
| smoser | ... | 23:48 |
| smoser | oh man. | 23:48 |
| smoser | that is just a mess. | 23:48 |
| smoser | (not your code, but the whole non-deterministic network state behavior of cloud-init) | 23:48 |
| smoser | magic fixes in 2.0 | 23:48 |
| smoser | i have to read/think more than i hve time for now. will try to look tomorrow | 23:49 |
| suro-patz | sure smoser | 23:49 |
| smoser | later. | 23:50 |
| smoser | and happy st. patricks day :) | 23:50 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!