/srv/irclogs.ubuntu.com/2021/07/15/#cloud-init.txt

blackboxswparide: thanks for the review on https://github.com/canonical/pycloudlib/pull/158 for lxd ephemeral shutdown handling I've pushed changes to also capture that ephemeral state at instance launch time17:28
ubottuPull 158 in canonical/pycloudlib "Lxd shutdown no wait" [Open]17:28
blackboxswbut I don't think we can drop the full ephemeral logic to check for instance type from "lxc info" because of supporting cloud.get_instance() calls17:29
blackboxswrharper: falcojr if either of you get a chance the I've pushed a couple of commits to the hotplug branch to better support rpm builds on centos/rocky. Things look really good on Ubuuntu. But, I still think there is still some work for CentOS/Rocky/RHEL for us to handle in followup PRs:17:34
blackboxsw1. Centos/Rocky/RHEL SElinux enforcing systems  need a custom policy for the cloud-init-hotplugd.socket|service behavior and FIFOs https://bugs.launchpad.net/cloud-init/+bug/193622917:35
ubottuLaunchpad bug 1936229 in cloud-init "selinux cloud-init-hotplugd.socket not having permissions to fifo sockets" [Undecided, New]17:35
blackboxsw2. Centos/Rocky/RHEL: sysconfig and NetworkManager collisions, cloud-init writes NM_CONTROLLED=no by default to ifcfg-eth1, so our net/activators.py trying to bring_(up|down)_interface will either collide with NetMgr behavior because it didn't reload the ifcfg-eth1 file or "nmcli connection up eth1" will claim no such managed device found.  17:39
blackboxsw3. *SuSE: might need to instrument a wicked net/activator.py for better support of interface bring up/down during hotplug17:40
blackboxswagain these issues I feel need to be addressed as separate PRs after the current one https://github.com/canonical/cloud-init/pull/936 as that pr is functional in it's own right for netplan-based systems17:42
ubottuPull 936 in canonical/cloud-init "Initial hotplug support (SC-19)" [Open]17:42

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!