juju bootstrap of a manual cloud erases dns configuration from machine
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Canonical Juju |
Triaged
|
High
|
Unassigned |
Bug Description
Setup :
With terraform, I created vms in Azure. Them, I ssh into one of them and bootstrap a manual cloud.
The problem is that juju wipes the dns configuration in systemd-resolve. Before the bootstrap action, my systemd-resolve status shows a list of 4 dns servers, as well as a DNS domain. After launching the bootstrap, that information is removed from systemd-resolve and the bootstrap hangs at "Running machine configuration script...".
While it was hanging there, I confirmed in another terminal that the dns info was gone and that my vm was not able to resolve anythin anymore. A simple sudo apt update returns resolution errors.
To work around this, while the bootstrap was hanging, I set the dns back to its original value with systemd-resolve --set-dns=
continue normally.
I can confirm that this happens with juju 2.8.6/stable and 2.8.5 on bionic vms (ubuntu-
tags: | added: cpe-onsite |
description: | updated |
description: | updated |
Changed in juju: | |
milestone: | 2.8.8 → 2.8.9 |
Changed in juju: | |
milestone: | 2.8.9 → 2.8.10 |
Changed in juju: | |
milestone: | 2.8.10 → 2.8.11 |
juju doesn't directly change the systemd-resolve configuration as far as I am aware.
It does use an upstream library github. com/coreos/ go-systemd to create/update service config for the juju agents and mongo.
More investigation needed to understand what is messing with the systemd resolve config.