raidandfade | howdy, me again! can't for the life of me get centos8 to learn network settings from cloud-init - i can see it reading network settings from openstack and it even has the sysconfig/ifcfg configs (apart from accept_ra which is false despite being set to true) - but the interface is still just dhcp.. additionally, there's ens3 interfaces lying around | 00:02 |
---|---|---|
raidandfade | that dont seem to be doing anything? | 00:02 |
minimal | raidandfade: have you tried enabling debugging for cloud-init so it will log more detail of what it is doing? | 01:07 |
raidandfade | I haven't, admittedly. I had the same sort of behavior on centos7 as well but i worked around it by adding a bootcmd to restart the network service after cloud-init completed | 01:07 |
raidandfade | same thing doesnt seem to work here | 01:08 |
minimal | if you edit /etc/cloud/cloud.cfg.d/05)logging.conf there are several "level=INFO" entries that you can change to "level=DEBUG" and then reboot | 01:12 |
raidandfade | Stderr: Error: Nexthop has invalid gateway. - fun | 01:14 |
raidandfade | looks like its not adding static routes with link scope? | 01:14 |
raidandfade | i can't find any explanation of why ipv6 isn't being added though, or why it converts accept_ra to IPV6_FORCE_ACCEPT_RA=no | 01:16 |
raidandfade | anything specific you want out of the log? im doing a search for "network" and it looks pretty similar to what i get out of debian with debugging | 01:17 |
raidandfade | all the net configs look like what i expect them to be, and there's no errors when writing them either: util.py[DEBUG]: Writing to /etc/sysconfig/network-scripts/ifcfg-eth0 - wb: [644] 272 bytes | 01:19 |
minimal | raidandface: I've only used static IPv6 so far myself | 01:19 |
minimal | are you using v1 or v2 network YAML format? I tend to use v2 | 01:20 |
raidandfade | the ip address is static, it's just accept_ra that is anomalous in our setup | 01:20 |
raidandfade | centos8 only supports sysconfig as far as i can tell, so v1? | 01:21 |
raidandfade | the datasource is openstack | 01:21 |
minimal | I'm using eni with NoCloud | 01:21 |
raidandfade | ill give it a shot with eni | 01:21 |
minimal | think only Debian and Alpine (and possibly Ubuntu) support eni | 01:30 |
raidandfade | yeah thats what i was thinking as well | 01:30 |
raidandfade | eni is effectively ifupdown, which is replaced by networkmanager (nmcli) on centos8 | 01:30 |
raidandfade | oddly enough... rebooting the system fixed all the network issues? | 01:31 |
raidandfade | it's accepting RA and has all ips | 01:31 |
raidandfade | and the non-scoped routes from before have scopes now | 01:32 |
minimal | perhaps the issue was only on 1st boot? typically during that it will have created the sysconfig stuff and then subsequent boots will use that file directly | 01:35 |
raidandfade | odd | 01:39 |
=== EugenMayer5 is now known as EugenMayer | ||
hffgbbbffd | Anyone possibly know why all my sudo commands timeout from cloud init but not other services? I googled around and saw some bugs in systemd and selinux, but I'm on all those updates versions. | 16:14 |
holmanb | hffgbbbffd: what's your cloud-config? | 16:35 |
hffgbbbffd | holmanb https://pastebin.com/EZy91zrV | 17:13 |
hffgbbbffd | contents of /etc/cloud/cloud.cfg | 17:13 |
henk | https://cloudinit.readthedocs.io/en/latest/topics/format.html AFAICT »make-mime subcommand takes pairs of (filename, “text/” mime subtype) separated by a colon« is phrased incorrectly. it should be something like »The make-mime subcommand’s option "-a" can be given multiple times and takes a pair of 'filename' and '“text/” mime subtype' separated by a colon« | 23:25 |
henk | because the subcommand itself does not take pairs, it seems from the following example | 23:25 |
minimal | henk: open an PR in the cloud-init github repo or raise bug on launchpad? | 23:52 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!