[01:00] <akutz> falcojr: This is confirmed as a bug in v21.3 per this comment https://github.com/kubernetes-sigs/image-builder/issues/712#issuecomment-938257637
[01:00] <akutz> (includes a way to repro with a unit test)
[01:18] <akutz> falcojr: I've also identified a patch for the issue, and I'll submit a PR in the morning. The patch is at https://github.com/kubernetes-sigs/image-builder/issues/712#issuecomment-938262835.
[08:13] <otubo> blackboxsw, rhansen hey guys I didn't see any messages about cloud-init summit this year. You guys thinking about having one?
[14:28] <akutz> I filed https://github.com/canonical/cloud-init/pull/1058 to address https://bugs.launchpad.net/cloud-init/+bug/1946493
[15:14] <H__> how to try cloud-init 21.3 on Ubuntu 20.04.3 in AWS ? (I see only 20.04.2 there)
[15:38] <manuvakery1_> holmanb: yes i am using the same Openstack version.  the Local boot stage is where I am getting the error because the known_mac dict contains only {'00:00:00:00:00:00': 'lo'} 
[17:36] <holmanb> manuvakery1_: net.get_interfaces() filters out various interface types (vlan/bond/bridge/blacklisted driver/etc) does anything there stand out to you as a possibility?
[17:37] <holmanb> If not the best way forward might be to just file a bug (including cloud-init collect-logs output and as much system detail as you can manage)
[17:38] <blackboxsw> otubo: I think we were going to delaying until Mar 2022, but wanted to float that idea by the mailinglist to get concensus on timing.
[17:39] <manuvakery1_> holmanb: I will bebug more and update
[17:39] <blackboxsw> *consensus
[18:39] <akutz> falcojr: Thanks for the quick merge1
[18:39] <akutz> Er, merge! :) 
[18:48] <minimal> akutz: I've been using v2 file with set-name for many months now but not with per-interface nameservers, rather using global nameservers so guess its by luck I didn't spot this previously
[18:50] <akutz> Heh. Well, Cluster API Provider for vSphere (CAPV) uses "set-name" to "normalize" interface names. It just so happened this person *also* added the DNS entry feature. When I was running CAPV I basically modeled its guest network config on Cloud-Init's v2 model :) https://github.com/kubernetes-sigs/cluster-api-provider-vsphere/blob/66795d613fe06b4ff7e8f4fe3beb63c4d53c9d4d/api/v1alpha4/types.go#L186-L279
[18:51] <akutz> Here's a Golang Template for transforming the above data into a valid Cloud-Init v2 network config :) https://github.com/kubernetes-sigs/cluster-api-provider-vsphere/blob/66795d613fe06b4ff7e8f4fe3beb63c4d53c9d4d/pkg/util/constants.go#L19-L89
[18:52] <akutz> minimal: But I never caught this either, because I was also always using global nameservers, so what can ya do? I'm just glad I had the epiphany about the set-name. That seemed suspicious to me since there was a KeyError.
[18:53] <minimal> I guess if you define per-interface nameservers for more than 1 interface that its supposed to use resolvconf to handle this?
[18:56] <akutz> I mean, I've no clue. Looking at netplan I think they expect one interface is the default route and you set them there? https://netplan.io/reference/#common-properties-for-all-device-types
[19:01] <Nico1> Hey just wanted some guidance on this error
[19:02] <Nico1> this morning after reboot clud-init gave: azure.py[ERROR]: Could not crawl Azure metadata: 'Dat
[19:02] <Nico1> aSourceAzure' object has no attribute 'failed_desired_api_version'
[19:03] <Nico1> after that it reconfigured everything as it it were firstboot
[20:01] <blackboxsw> Nico1: can you run `cloud-init --version` 
[20:03] <Nico1> sure /usr/bin/cloud-init 21.3-1-g6803368d-0ubuntu1~18.04.3
[20:08] <blackboxsw> thanks Nico1:  Can you please file a bug at  https://bugs.launchpad.net/cloud-init/+filebug  and attach the tar.gz from `cloud-init collect-logs to help triage here? Ok so you are ubuntu bionic Azure. That attribute was added in 21.1-19 which was released in March I think.
[20:08] <blackboxsw> I'm guessing you can from a version earlier that 21.1-19 straight to 21.3.1, but your /var/log/cloud-init.log will tell us better
[21:14] <Nico1> will do thanks, on the other hand I validated that package was updated on october, but from v21.2