UbuntuIRC / 2020 /09 /30 /#cloud-init.txt
niansa
Initial commit
4aa5fce
[12:02] <meena> trying to change my name in launchpad vs ubuntuone is a pain
=== vrubiolo1 is now known as vrubiolo
[12:43] <meena> https://github.com/canonical/cloud-init/pull/588 ⬅️ finally started
[13:03] <Odd_Bloke> \o/
[13:04] <Odd_Bloke> powersj: Thanks for the background!
[13:31] <Odd_Bloke> Oh right, and I can't even hack around it by specifying a key because my key has a password and paramiko doesn't know how to handle that.
[13:31] <Odd_Bloke> So for my case, I _need_ to use my SSH agent.
[16:42] <Odd_Bloke> blackboxsw: https://github.com/canonical/pycloudlib/pull/34 is my first PR for addressing the "cloud-init failures unconditionally raise exception" issue; it's a refactor that will be followed by the functional change. You touched this code most recently (and the most), do you have time for a review today?
[16:45] <Odd_Bloke> (I'm fixing the CI failure now; accidentally used a new-in-3.8 mock feature.)
[16:59] <meena> what's pycloudlib?
[17:45] <Odd_Bloke> meena: The library that the new cloud-init integration testing is being built on; also used for another integration testing framework on our team.
[18:16] <Odd_Bloke> falcojr: blackboxsw: OK, my pycloudlib yak shaving has now lead me to https://github.com/canonical/pycloudlib/issues/35
[18:17] <Odd_Bloke> Thoughts on that proposed refactor would be much appreciated.
[18:44] <meena> i think i'm mostly asking, cuz the name reminds me of… https://libcloud.apache.org/
[18:56] <Odd_Bloke> Yeah, we considered libcloud, but (IIRC) it didn't have support for image manipulation (i.e. snapshotting an instance as an image, after we've installed the cloud-init we want to test in it), and we also had concerns about being able to test cloud-specific features (which necessarily won't be surfaced in an API designed to work for many clouds).
[18:59] <Odd_Bloke> Aha, I've found our internal doc from when pycloudlib was being created: it was hard-coding regions and instance types (which as we will often be asked to test new regions or instance types, would mean we would be semi-regularly having to maintain a fork/hack things), and it wasn't able to create a VPC on EC2.
[21:01] <Odd_Bloke> falcojr: I'm looking at a tox.ini environment for the tests, and perhaps Travis integration, FYI.
[21:02] <falcojr> Odd_Bloke: looking at how?
[21:03] <Odd_Bloke> Fair point, more "writing" than "looking". :p
[21:52] <falcojr> ok, I'm touching the tox.ini for logging and default capture reasons too
[21:57] <falcojr> also, just found this pytest ini file option...
[21:57] <falcojr> disable_test_id_escaping_and_forfeit_all_rights_to_community_support (bool):
[21:57] <falcojr> :D