* dilfridge gives up further attempts for today ("clearly I'm holding it the wrong way up") | 00:02 | |
minimal | dilfridge: as an Alpine user I don't use cloud-init with systemd so I'm unfamiliar with the services | 00:30 |
---|---|---|
minimal | there was some systemd "simplification" as part of #5818 and #5830 | 00:31 |
holmanb | dilfridge: lmk if you have questions after reading 5489 | 00:59 |
holmanb | The breaking changes doc page has some more info regarding that change | 00:59 |
holmanb | And the blog post should help too, which I can link if you can't find | 01:00 |
holmanb | dilfridge: but yeah enabling that service isn't the expected way to enable cloud-init | 01:01 |
holmanb | it's not very well documented, but there is a systemd generator which should automatically enable cloud-init | 01:04 |
holmanb | It uses ds-identify to identify which cloud it's running on | 01:04 |
holmanb | (and if it can identify a cloud at all) | 01:05 |
holmanb | If it identifies a cloud, then it adds cloud-init.target to the systemd boot transaction | 01:05 |
holmanb | and the other services get pulled in via respective WantedBy=cloud-init.target | 01:06 |
holmanb | This allows a single image to launch on various different clouds and run the correct logic. | 01:09 |
holmanb | Alternatively if you want your image to be tailored to a specific cloud you could set a single datasource | 01:10 |
holmanb | all of this is true before and after #5489, hopefully this context helps | 01:13 |
holmanb | dilfirdge: not sure about the broken link, would need more context | 01:15 |
holmanb | dilfridge: oh, and cloud-init.service was renamed to cloud-init-network.service for exactly the reason that you stumbled on | 01:19 |
holmanb | it's perfectly reasonable for a person to assume that enabling/starting cloud-init.service is how to interact with the cloud-init service - but that couldn't be further from the truth | 01:19 |
holmanb | cloud-init.service is the second of four services, that run at different times - nothing more, nothing less | 01:20 |
holmanb | s/is/was/ | 01:20 |
holmanb | cloud-init.service was a terribly confusing name for the second stage | 01:42 |
holmanb | there are reasons why it was named that way but those reasons are just a historical footnote today | 01:42 |
holmanb | The change we made is painful, but hopefully for the better in the long run | 01:44 |
holmanb | dilfridge: recommended reading -> https://cloudinit.readthedocs.io/en/latest/explanation/boot.html | 01:44 |
holmanb | dilfridge: we should talk about systemd ordering for gentoo too when you get a chance | 01:54 |
holmanb | But I'm off for the night | 01:54 |
dilfridge | no offense and sorry if I'm asking stupid stuff, but ... even the boot ^ description page does not answer - what initiates the first step? i.e. what runs ds-identify ? | 13:22 |
dilfridge | (hmm, apparently the systemd generator does that...) | 13:28 |
dilfridge | hmm the problem seems to be that ds-identify does not find the cdrom image for NoCloud | 13:32 |
dilfridge | no, it does find the iso - "ISO9660DEVS=/dev/sr0=CIDATA" and "Found single datasource: NoCloud" | 13:49 |
dilfridge | ls | 13:49 |
dilfridge | oh LOL... " ... line 1: netcat: command not found" | 13:51 |
* dilfridge adds netcat to runtime dependencies | 13:51 | |
holmanb | dilfridge: specifically net-analyzer/openbsd-netcat | 14:20 |
holmanb | Because unix socket support | 14:21 |
holmanb | dilfridge: no offense taken :) | 14:22 |
holmanb | Okay I see you figured that out already | 14:27 |
dilfridge | https://github.com/canonical/cloud-init/pull/5933 | 14:30 |
-ubottu:#cloud-init- Pull 5933 in canonical/cloud-init "fix: use program name of netcat as installed by upstream, 'nc'" [Open] | 14:30 | |
dilfridge | netcat installs itself by default only as "nc", the long name is debian/ubuntu special | 14:30 |
dilfridge | and now it works again :) | 14:31 |
holmanb | At least one other distro uses it - arch I think? | 14:33 |
holmanb | But if nc is more standard then of course we should use that | 14:34 |
holmanb | Thanks for the PR! | 14:34 |
holmanb | Probably won't get to it today | 14:35 |
dilfridge | I checked our build system and it doesn't pass any specific binary name, and I checked the ubuntu control file, where "netcat" is added as slave under alternatives | 14:35 |
dilfridge | no hurry | 14:35 |
holmanb | Glad you got it working again | 14:35 |
dilfridge | learning a lot here :D | 14:35 |
holmanb | Heh, yep :) | 14:36 |
holmanb | And if the docs are lacking PRs and reports are always welcome. | 14:37 |
holmanb | The docs have come a long way in the last couple of years but they still have a long way to go | 14:38 |
dilfridge | the next image build is in a few hours, then with 24.4 and the patch, looking forward to trying that out | 14:40 |
holmanb | Nice | 14:41 |
holmanb | There's also a dependency that was dropped in the last year-ish: python3-netifaces | 14:42 |
holmanb | I don't remember if we got that updated in gentoo or not | 14:42 |
holmanb | And I don't recall if gentoo had any distro-dependent tests disabled, but if so those can probably be dropped | 14:44 |
holmanb | Because we cleaned up some distro dependant test failures and added an alpine-based unit test job to ci | 14:45 |
dilfridge | I'll have a look, sure | 15:07 |
dilfridge | looks like nothing creates the dir /run/cloud-init/share yet | 23:25 |
* dilfridge needs to verify that still, but tomorrow | 23:25 | |
dilfridge | (that makes the generation of the sockets fail with "No such file or directory" in every service) | 23:31 |
dilfridge | root | 23:39 |
dilfridge | oops | 23:39 |
dilfridge | also, do the 4 services have to be enabled "manually" or is the generator somehow supposed to take care of that? :| | 23:40 |
holmanb | Generator does that | 23:55 |
holmanb | like I said, the generator enables cloud-init.target | 23:56 |
holmanb | (it just manually creates the symlink) | 23:56 |
holmanb | And the other services have WantedBy=cloud-init.target in the [Install] section | 23:57 |
holmanb | Err, /run/cloud-init/share should be generated automatically. | 23:59 |
holmanb | Pastbin a log? | 23:59 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!