[08:28] <testing1818282> Hello Odd_Bloke hope you remember me from yesterday. I had a question regarding ssh host key keys that I think get generated too late. I am sorry for leaving in the middle of the conversation but my isp has failed on me and couldn't log back in. Do you have any idea what I can do to have my machine print the keys? Thanks for the help
[08:31] <testing1818282> These are the logs I have on hand: https://pastebin.ubuntu.com/p/DrZ84SWsfM/ & https://pastebin.ubuntu.com/p/dk4hMqsCVX/
[16:05] <beantaxi_> Hi all ... thought I'd ask about how PS #745 is looking (https://github.com/canonical/cloud-init/pull/745). If there's anything needed from me to help move it forward I'm happy to help. Thanks!
[16:13] <rharper> Odd_Bloke: in case testing1818282 returns, it looks to me like, the instance 1) already has existing host keys; they get deleted in the log 2) the instnace does not invoke ssh-keygen, which could be cause of user-data config , or the RHEL image it's running may generate ssh-keys via the ssh-keygen.service; in which case maybe that is running (or not running) at the correct time;
[16:15] <rharper> beantaxi: it's on my list to review;  my quick take is that each of those scripts are 99% identical other than the name and file;  I would like to see a common handle implementation that each of those script files import and use, passing in the variable name/path;
[16:33] <Odd_Bloke> rharper: Aha, didn't know about that service, that's really handy info.  We'll see if I get to use it. :p
[16:37] <Odd_Bloke> https://github.com/canonical/cloud-init/pull/750 is open, clarifying how netplan passthrough works; I couldn't find anywhere that we documented it previously.
[16:44] <rharper> Odd_Bloke: thanks for 750, I'll review;  I myself have had to re-read the code a few times to recall how we keep the imported config around and detect v2 pass-through to write out the original
[16:45] <minimal> Odd_Bloke: that PR came about from a discussion I'd had with the reporter on here over the holiday period
[16:46] <minimal> I was pointing out to him that just because what he was doing with netplan currently worked in c-i was no guarantee that it would continue to work as it wasn't a documented c-i network-v2 configuration
[16:46] <minimal> plus the issue of course of it not working for e/n/i and other renders
[16:49] <Odd_Bloke> minimal: Aha, right, I hadn't put the IRC backlog and bug together; thanks for helping them out!
[16:50] <minimal> at the time it surprises me to see him using the "from:" value as I've used static routes with e/n/i and had not seen that before
[16:51] <Odd_Bloke> rharper: I've just seen https://paste.ubuntu.com/p/CWcp88S2QM/ in a focal cloud-init KVM run; does that look like the same kernel issue we've been seeing from time-to-time?
[16:52] <rharper> Odd_Bloke: that looks new to me;
[16:54] <rharper> the stack track looks KVM related;  I would guess either the host OOMed or otherwise couldn't fill the memory request; so highly memory contention;
[16:59] <rharper> Odd_Bloke: looks like this, https://github.com/torvalds/linux/commit/2a266f23550be997d783f27e704b9b40c4010292  ; which landed in 5.11, not sure if kernel team has that in 5.4.1 in focal ;
[17:07] <beantaxi> rharper: Sounds reasonable. I can do that today. Thanks for the feedback!
[17:07] <rharper> beantaxi: cool!
[17:08] <rharper> beantaxi: I also elaborated on the PR directly, please look at those comments as well
[17:20] <Odd_Bloke> rharper: That looks like it landed in v4.15, if I'm reading the GH tags right?
[17:35] <rharper> Odd_Bloke: yeah, I don't fullly understand what that 4.15 tag means;  if it did land in 4.15, how did it get into 5.4, and why also 5.11
[17:36] <Odd_Bloke> I believe that displays the tags which include this commit in their history; so I think it landed in 4.15 (and is therefore present in every subsequent tag).
[17:42] <Odd_Bloke> We're also seeing https://paste.ubuntu.com/p/N395dZxnGz/ on groovy.
[17:57] <beantaxi> rharper: Yes I read through your comments and smosers too. They're great: politeful and respectful but very clear. Much appreciated!
=== ijohnson is now known as ijohnson|lunch
[19:17] <Odd_Bloke> rharper: Thanks for the review. :)
=== vrubiolo1 is now known as vrubiolo
=== ijohnson|lunch is now known as ijohnson
[22:02] <wyoung> o/
[22:04] <wyoung> I am using cloud-init with AWS EC2 deployed via CloudFormation.  If I change the cloud-config can I get it to apply on an CloudFormation update?  For example, I want to update the contents of a file created via write_files.
[22:05] <rharper> wyoung: on an existing instance? or new instance?
[22:06] <wyoung> rharper: existing.
[22:07] <rharper> by default, cloud-init won't re-run write_files config on the same instance;  you can force a particular module to run with:  cloud-init single --name write_files --frequency always ;
[22:08] <wyoung> rharper: I would add that to the runcmd section?
[22:08] <wyoung> and run it manually for the existing instance?
[22:12] <rharper> you need to run that command on the instance itself  (ie, ssh into and run that command)
[22:12] <rharper> wyoung: ^
[22:13] <wyoung> rharper: And for new instances (after I add in that command to runcmd) will it write files on a change?
[22:13] <wyoung> Or do I need to manually run the command again after any change/.
[22:13] <rharper> not really; manually need to run that any time it changes;
[22:15] <rharper> I'm not sure how CloudFormation works; but ideally; you could set up some http long poll against the CF config URL; and on-change, run an update; but that's just not in place;  any given cloud-init config moodules (run_cmd, users, packages, etc...) may not be idempotent; so it's not clear what cloud-init should do if the user-data config changed ...
[22:16] <rharper> for something like a write_files where you know you want to update a particular file, you may want to set up some sort of watch on the config URL, and on change run the command;
[22:16] <rharper> less effort but not as dynamic would be put that into a cron job which runs every so often