=== cpaelzer__ is now known as cpaelzer | ||
otubo | smoser: IIUC you want cloud-init --version to report the full name version (like 20.3-2-g371b392c-0ubuntu1~20.04.1) instead of just 20.3? | 12:13 |
---|---|---|
otubo | smoser: regarding https://bugzilla.redhat.com/show_bug.cgi?id=1897616 - we don't normally patch rhel7 anymore unless it's critical or a CVE. Not sure how much people will touch this corner case since RHEL always ships (and uses) NetworkManager | 12:17 |
ubot5 | bugzilla.redhat.com bug 1897616 in cloud-init "cloud-final.service fails if NetworkManager not installed." [Medium,New] | 12:17 |
smoser | otubo: wrt 1897616, that is a reasonable response, but the change that caused the regression was recent (last month or so ? whenever that change to 19.4 came in. but thats fine). | 13:18 |
smoser | but on rhel 8, it seems reasonable to *not* fail in an otherwise working scenario | 13:19 |
smoser | sysconfig *does* work, and the change from upstream breaks it. | 13:20 |
smoser | now, wrt cloud-init --version. yeah, i'd like to see cloud-init --version put 20.3-7.el8 | 13:20 |
smoser | or whatever is appropriate there. | 13:20 |
smoser | otubo: and again, thank you for getting back to me | 13:28 |
otubo | smoser: agreed to patch it for >= RHEL-8*. Will clone that BZ and apply your patch. | 13:34 |
otubo | smoser: thank you for catching that :) | 13:35 |
smoser | otubo: is there a *real fix* ? | 13:37 |
smoser | that doesn't require basically doing a "restart" ? | 13:37 |
smoser | as that'd be ideal. | 13:37 |
smoser | rharper offered that you could actually put a dropin to cloud-init-final.d in the NetworkManager package. | 13:38 |
otubo | smoser: I understood the dropin would be on cloud-init, to restart NM at the end of cloud-init-service, like I did. | 13:51 |
otubo | (had to read the logs again to remember) | 13:51 |
otubo | rharper said: https://paste.ubuntu.com/p/tGyGWJBq5t/ - that is a systemd unit drop it config which would trigger the try-restart right before cloud-init.service | 13:52 |
otubo | Unless I'm missing something. | 13:52 |
otubo | smoser: the real ideal fix would be to make NM to wait until cloud-init really finishes to write all its net config before starting. | 13:52 |
smoser | otubo: well, it somewhat makes more sense to put the dropin in NM | 14:55 |
smoser | as it is NM that is modifying behavior. | 14:55 |
smoser | rather than adjusting cloud-init's behavior wrt a package that might not be installed. | 14:56 |
smoser | that way the modification only comes if the thing is installed. | 14:56 |
smoser | but it doesn't really matte.r.. ideally we fix what is actually wrong | 14:56 |
smoser | rather than patching to restart. | 14:56 |
rharper | otubo: the reason for moving the drop-in to Network-Manager is a packaging one; if your image does not contain NM, but it has cloud-init; then the drop-in as-written fails since the systemctl command fails if the service does not exist; | 15:12 |
otubo | rharper: smoser I think it makes sense. But I guess Scott's patches are ok for now. It's much easier to squeeze in a small patch like that for cloud-init right now. I'll probably file a new BZ for NetworkManager later (RHEL-8.4, perhaps) | 15:18 |
rharper | otubo: ok; I think it would be good for 8.x bug to dig deeper and to figure out the ordering issue; either with a further change to cloud-init in the sysconfig path, or some reliable way to delay or restart NM parsing of config files once cloud-init has finished writing complete config; | 15:25 |
minimal | general question: any specific reason why the various /etc/hosts template files (templates/hosts.*.tmpl) all list localhost.localdomain before localhost? It means that running a netstat on a machine shows the long-form localhost name rather than the short-form name. | 16:01 |
minimal | same question applies to short/long localhost6 and hostname/fqdn | 16:01 |
Odd_Bloke | I don't know of a particular reason; the Debian template doesn't (except with hostname/FQDN), which suggests to me that that behaviour originates from Red Hat: https://github.com/canonical/cloud-init/commit/09f6384ac5694be18bef4872c9f19b9601b48f8b | 16:59 |
minimal | Odd_Bloke: was wondering if it was worth me raising a MR to "fix" it. | 17:03 |
minimal | I'm certainly surprised to see the long-form localhost and localhost6 used by default | 17:04 |
minimal | the Debian template file doesn't have the long-form of localhost at all, rather than the order being in question | 17:07 |
Odd_Bloke | Right, I meant (but didn't say :p) that as this behaviour isn't present on Debian/Ubuntu, I wouldn't rely on my (lack of) knowledge about it: it might be totally correct on Red Hat for reasons I'm not aware of. | 17:20 |
meena | i realise that this pr https://github.com/canonical/cloud-init/pull/655 only has a review from people who don't make green ✅ but i'd love to see this merged in the current release, so we don't have to carry around the patches in the ports | 17:43 |
meena | or has that window of opportunity closed? | 17:45 |
=== ijohnson is now known as ijohnson|lunch | ||
meena | the deafening silence could be a hint, but I've been so under water the past week or two that i haven't been very good at noticing things | 18:27 |
powersj | Odd_Bloke, blackboxsw rick_h is there talk about 20.4 soon? Can we also get the topic updated for office hours? I don't recall if that was a discourse thing or not... | 19:37 |
=== ijohnson|lunch is now known as ijohnson | ||
=== MAbeeTT_ is now known as MAbeeTT |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!