| Ally_ | Hello, hope everyone is well. I don't suppose anyone can advise if there's a way to set the order of packages to be installed when specifying them in the YAML custom script? I want to install 'ansible' but I need to install 'oracle-epel-release-el7' first, putting it first still seems to get ignored, I imagine it does it alphabetically or similar? | 12:28 |
|---|---|---|
| Ally_ | Or, should I just do it with a runcmd instead? | 12:29 |
| aciba | Ally: There is an undocumented (atm) way to do that using https://cloudinit.readthedocs.io/en/latest/topics/merging.html. Maybe falcojr could help you. | 12:42 |
| Ally_ | Thanks aciba, much appreciated! | 13:21 |
| falcojr | Ally_: when you say order of packages to be installed, are you referring to using the 'packages' key in cloud-init userdata? E.g., packages: ['oracle-epel-release-el7', 'ansible'] | 14:33 |
| falcojr | or something else? | 14:33 |
| Ally_ | Hi falcojr, yep | 15:19 |
| falcojr | Ally_: that behavior is distro-dependent. Which distro? | 15:20 |
| Ally_ | We're using OEL 7.9 | 15:20 |
| Ally_ | I've worked around it by doing it via runcmd, but I assumed there was a 'smarter' way by using the native package command | 15:22 |
| falcojr | hmmm, do you know if that gets identified as RHEL by cloud-init, or do you have some custom code for identifying that? | 15:22 |
| Ally_ | It should identify as RHEL (it's a clone) | 15:22 |
| Ally_ | Is there a quick way to determine how cloud-init sees it? | 15:23 |
| blackboxsw | Ally_: I think cloud-init query distro | 15:24 |
| falcojr | blackboxsw: TIL! I would've said search the logs | 15:25 |
| Ally_ | thanks blackboxsw | 15:25 |
| blackboxsw | otherwise grep "Using distro class" /var/;log/cloud-init.log would cofirm | 15:25 |
| blackboxsw | right | 15:25 |
| Ally_ | # cloud-init query distro | 15:25 |
| Ally_ | ERROR: Undefined instance-data key distro | 15:25 |
| blackboxsw | ahh OEL 7.9 ... older cloud-init based on version 19.3 or something right | 15:26 |
| Ally_ | # grep "Using distro class" //var/log/cloud-init.log | 15:26 |
| Ally_ | 2022-09-09 13:19:43,979 - stages.py[DEBUG]: Using distro class <class 'cloudinit.distros.rhel.Distro'> | 15:26 |
| blackboxsw | cool that worksthx | 15:26 |
| Ally_ | Cloud-init v. 19.4 | 15:26 |
| falcojr | Ally_: so actually, it looks like regardless of distro it orders them in the way they were specified. But it does it with a single command. Are you looking for multiple, independent commands for installation? | 15:27 |
| Ally_ | Umm, no, not really. To install 'ansible' package it needs the EPEL repo enabled, and that's an RPM install on OEL. So I specified the epel package first in the list of packages and 'ansible' second, but when it ran it tried to install the 'ansible' package first. | 15:29 |
| Ally_ | https://pastebin.com/UVrFbJFj | 15:31 |
| Ally_ | That's what I was trying to get to work | 15:31 |
| Ally_ | I kinda assumed it was doing like a hoist in JS and running things alphabetically or something | 15:33 |
| waldi | Ally_: you are looking for yum_repos, which asks cloud-init to add repos | 15:34 |
| waldi | cloud-init will of course not install packages one after one, but everything in one go | 15:34 |
| Ally_ | Ah, ok, I think I see what you mean. So if I enable the EPEL using the yum_repos command I can skip the epel package part altogether. | 15:35 |
| waldi | yes | 15:35 |
| Ally_ | I will test that now, thank you waldi (and everyone else!) | 15:35 |
| blackboxsw | note Ally_ you are specifying `ansible:` config in cloud-init but you are running cloud-init version 19.4 on OEL.... `ansible:` support in cloud-init was only added after cloud-init version 22.2 so you'll probably have to use individual `runcmd:`declarations to perform that setup of ansible in your distro. | 16:12 |
| blackboxsw | as I recall, python version on OEL is python2.7 right? which means that OEL systems won't be able to install and run anything beyond 19.4 unless they also happen to have python3 configured | 16:13 |
| Ally_ | Ah, that's good to know, thank you! | 16:18 |
| blackboxsw | note we have a copr build repo https://copr.fedorainfracloud.org/coprs/g/cloud-init/el-testing/ that may help wih EL8-based images if you have them | 16:24 |
| blackboxsw | it contains cloud-init 22.3 | 16:24 |
| cjp256 | blackboxsw/falcojr: any thoughts what may be happening here? https://pastebin.ubuntu.com/p/kDc5ZKRZVs/ | 17:57 |
| cjp256 | holman^ | 17:57 |
| cjp256 | (search for first hit of iid-datasource) | 17:58 |
| blackboxsw | looking | 18:01 |
| cjp256 | get_data() is also called twice, init-local and init | 18:02 |
| blackboxsw | strange mid-init stage iid-datasource shows up | 18:03 |
| cjp256 | this is on upgrade from 22.2 -> 22.3 in proposed | 18:05 |
| cjp256 | thought maybe it was that recent cache change but I reverted it and didn't seem to change that | 18:05 |
| blackboxsw | cjp256: so iid-datasource comes into play when cloud-init no longer sees "instance-id" in the cached metadata that was read | 18:07 |
| blackboxsw | I'd be curious if something between init-local and init stage blew away self.metadata on the obj.pkl | 18:08 |
| cjp256 | yeah, it should be there - i lost a debug log, i'll retry | 18:08 |
| * blackboxsw takes a break from internal work and starts plugging away at Azure upgrade path on -proposed to see what's up there | 18:38 | |
| falcojr | ^ I would normally jump in to help, but I'm guessing it's a welcome break :P | 18:40 |
| holman | cjp256: what's puzzling me is that you said building from tag 22.3.1 didn't cause the same issue | 18:40 |
| holman | otherwise I'd guess that the self._reported_ready_marker_file addition not being unpickled as a likely cause | 18:41 |
| cjp256 | I take that back, I think it may have been because I went from the proposed deb to the custom one? | 18:46 |
| cjp256 | i retried that today and it did repro | 18:46 |
| cjp256 | from a fresh 22.2 | 18:46 |
| holman | 22.2->22.3.1 repro'd the issue? | 18:47 |
| cjp256 | yeah the current image 22.2-0ubuntu1~22.04.3 | 18:47 |
| holman | gotcha | 18:47 |
| cjp256 | i'll resume looking into it monday, have a good weekend! :D | 18:48 |
| holman | yeah I'm pretty sure there's an issue on unpickle here, since _reported_ready_marker_file won't exist in 22.2 pickles, | 18:48 |
| holman | cjp256: You too! | 18:48 |
| falcojr | holman: we should figure out why the upgrade integration test didn't catch this. Do you need help digging into that? | 19:31 |
| holman | @falcojr: Saw your comment on the PR, I'm looking at that now. I should be more familiar with this and I've never touched it, so I'll reach out if I need help. | 19:32 |
| holman | @falcojr: looks like test_upgrade in SRU testing used an image_id: Canonical:0001-com-ubuntu-server-jammy-daily:22_04-daily-lts | 19:47 |
| holman | Daily images would have compatible pickles I think. | 19:47 |
| falcojr | holman: it's not the daily ppa. It's the daily cloud image. I wouldn't expect it to launch with that property, so it shouldn't be in the pickle at upgrade | 19:50 |
| falcojr | i.e., I would expect Canonical:0001-com-ubuntu-server-jammy-daily:22_04-daily-lts to be providing 22.2, then the upgrade test upgrades to 22.3 proposed | 19:51 |
| holman | falcojr: I see, thanks. | 19:57 |
| holman | tangentially related, I see that the ubuntu cloud-image locator website isn't aware of the existence of Jammy images. | 19:58 |
| * holman goes back to digging into the test | 19:59 | |
| holman | s/Jammy images/Jammy images on Azure/g | 19:59 |
| CorvetteZR1 | hello. how do i edit an existing cloud-init config? | 21:26 |
| CorvetteZR1 | i set up a mount with noexec option. i removed it manually from fstab and remount; but when i reboot the server, i see noexec in fstab again. how can i modify it in cloudinit, so when it overwrites fstab, it no longer applies noexec ? | 21:27 |
| CorvetteZR1 | without reprovisioning whole server i mean | 21:27 |
| holman | @falcojr - not sure if you're still around, but I did a brain-dump of my working theory in that PR. | 21:59 |
| falcojr | holman: cool. I'll take a look (maybe Monday) | 22:00 |
| holman | @falcojr sweet, no rush | 22:00 |
| holman | CorvetteZR1: What's your config look like? cc_mounts runs once per instance, so running every boot is not expected. More information in general about your setup would help a lot :) | 22:02 |
| blackboxsw | holman: you are totally right on cjp256's Azure upgrade bug. it is unpickling problem across upgrade given that the obj.pkl doesn't have the _. the following patch fixes it https://paste.ubuntu.com/p/4yybCvHgvp/ | 23:31 |
| blackboxsw | I'll put up a PR and set you both as reviewers | 23:32 |
| blackboxsw | this will block SRU | 23:32 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!