| opendevreview | Ian Wienand proposed openstack/diskimage-builder master: containerfile: update test to jammy https://review.opendev.org/c/openstack/diskimage-builder/+/838981 | 04:50 |
|---|---|---|
| opendevreview | Maksim Malchuk proposed openstack/diskimage-builder master: Revert "Fallback to persistent netifs names with systemd" https://review.opendev.org/c/openstack/diskimage-builder/+/838863 | 06:27 |
| opendevreview | Maksim Malchuk proposed openstack/diskimage-builder master: Revert "Fallback to persistent netifs names with systemd" https://review.opendev.org/c/openstack/diskimage-builder/+/838863 | 07:34 |
| ianw | the containerfile elements in the gate look broken due to a podman point release | 07:46 |
| ianw | i've filed https://github.com/containers/podman/issues/13968 | 07:46 |
| opendevreview | Maksim Malchuk proposed openstack/diskimage-builder master: Revert "Fallback to persistent netifs names with systemd" https://review.opendev.org/c/openstack/diskimage-builder/+/838863 | 08:34 |
| *** Guest2246 is now known as prometheanfire | 13:54 | |
| opendevreview | Clark Boylan proposed openstack/diskimage-builder master: DNM trying to verify dhcp-all-interfaces fails with bullseye https://review.opendev.org/c/openstack/diskimage-builder/+/839079 | 15:46 |
| opendevreview | Clark Boylan proposed openstack/diskimage-builder master: Trigger dhcp-all-interfaces on udev move events https://review.opendev.org/c/openstack/diskimage-builder/+/839080 | 15:48 |
| clarkb | rpittau: ianw I left a comment on 838863 explaining these last two chagnes and proposing a path forward | 15:55 |
| clarkb | hopefully that all makes esnse? | 15:55 |
| opendevreview | Clark Boylan proposed openstack/diskimage-builder master: Trigger dhcp-all-interfaces on udev move events https://review.opendev.org/c/openstack/diskimage-builder/+/839080 | 18:10 |
| clarkb | doing this testing has exposed what might be a latent bug in the debuntu impleentation of dhcp-all-interfaces which that latest push attempts to address. Also, when I remove simple-init from the element list to prevent glean from running that means nothing is generating ssh host keys so ssh fails. But the tcp connections are coming in which makes me confused as to what was | 18:17 |
| clarkb | broken in the first place | 18:17 |
| clarkb | since as far as I can tell udev is firing for ens3, trying to run the script (failing due to the latent path bug which hpoefully is now fixed), then ssh can connect | 18:17 |
| clarkb | fungi: you're more familiar with eni than me can you make sense of why https://zuul.opendev.org/t/openstack/build/5fefdfecfd064facbca9d398045e8d51/log/instances/955decc1-7403-4f22-a31c-12a2237540b5/console.log#746 is complaining from this source: | 19:17 |
| clarkb | https://opendev.org/openstack/diskimage-builder/src/branch/master/diskimage_builder/elements/dhcp-all-interfaces/install.d/dhcp-all-interfaces.sh#L79-L94 | 19:17 |
| clarkb | also it looks like itgenerated ssh host keys in that console log so I'm not sure why it is failing. However it definitely apperas that the network is configured because we log that we failed to establish the layer 7 connetion to the lcient but l3 is there | 19:18 |
| fungi | clarkb: where can i find the /etc/network/interfaces file it's parsing there? | 19:19 |
| clarkb | fungi: we'd need to hold a node and copy the file out of the generated test vm image | 19:20 |
| fungi | my guess is that an iface line in there has five parameters, or at least the parser thinks it does | 19:21 |
| fungi | maybe some stray whitespace in one of the tokens? | 19:21 |
| fungi | hard to say without seeing what's actually on that line | 19:22 |
| clarkb | ok, I'm probably going to look at why ssh is failing there when networking seems to be ocnfigured | 19:23 |
| clarkb | on the client side it says root@172.24.5.59: Permission denied (publickey,password). which matches up with the failures in the server log | 19:23 |
| clarkb | But importantly I think this is working with predictable interface names? | 19:24 |
| clarkb | oh! I think we may be relying on the metadata server to supply the ssh key | 19:30 |
| clarkb | and without simple-init that breaks :/ | 19:31 |
| clarkb | let me see if I can just bake a key in | 19:31 |
| clarkb | fungi: the bad line 9 of /etc/network/interfaces is "iface ens3 inet6 autodhcp 0 | 21:34 |
| clarkb | except I think there is a tab between auto and dhcp? | 21:34 |
| clarkb | fungi: it is this line https://opendev.org/openstack/diskimage-builder/src/branch/master/diskimage_builder/elements/dhcp-all-interfaces/install.d/dhcp-all-interfaces.sh#L93 is it missing a \n like https://opendev.org/openstack/diskimage-builder/src/branch/master/diskimage_builder/elements/dhcp-all-interfaces/install.d/dhcp-all-interfaces.sh#L89 | 21:35 |
| fungi | interesting, yeah, i think for inet6 we want auto, for inet we want dhcp | 21:40 |
| fungi | and i have no idea what the 0 there is fort | 21:40 |
| fungi | for | 21:40 |
| clarkb | fungi: ya looking at the line aboe it seems they want it on a newline tabbed over with a 0 or 1 flag | 21:41 |
| clarkb | I'll try adding the missing \n | 21:41 |
| fungi | that seems likely, i concur | 21:41 |
| fungi | though not sure what the "dhcp 0" line is intended to do | 21:42 |
| fungi | Options: dhcp int: Use stateless DHCPv6 (0=off, 1=on) | 21:43 |
| fungi | okay, so that's a thing | 21:43 |
| fungi | per the interfaces(5) manpage | 21:43 |
| fungi | and yes, it needs to be on its own line | 21:43 |
| fungi | note that indentation is not semantic, it's just for readability | 21:44 |
| fungi | but newlines are important to the parser | 21:44 |
| opendevreview | Clark Boylan proposed openstack/diskimage-builder master: Trigger dhcp-all-interfaces on udev move events https://review.opendev.org/c/openstack/diskimage-builder/+/839080 | 21:44 |
| clarkb | ah | 21:44 |
| clarkb | ok cool so thats 2 real bugs in the debian implementation of dhcp-all-interfaces I think we've fixed so far :) | 21:45 |
| clarkb | that said I still havne't seen anything related to udev so not sure why we needed to break predictable interface names. Hopefully getting these jobs to pass will make people comfortable with the revert and these fixups | 21:45 |
| *** mazzy509881292958085949 is now known as mazzy50988129295808594 | 23:54 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!