deploying Ubuntu 18.04 with MAAS 1.9.5 fails during the final boot
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
MAAS |
Expired
|
Undecided
|
Unassigned | ||
curtin (Ubuntu) |
Invalid
|
Undecided
|
Unassigned | ||
Trusty |
New
|
Undecided
|
Unassigned |
Bug Description
I have a bunch of machines here enlisted in a MAAS 1.9.5 that commissioned happily. Deploying Ubuntu 16.04 to them works fine too, and they continue working when d-r-u'd to 18.04.
However, deploying 18.04 directly fails -- the nodes boot into the installer successfully and reboot at the end, but then when they boot into the final installation they fail to get network and MAAS eventually transitions the node to "Failed deployment". Here's the console log from such a final boot (Canonical-only link): https:/
I talked with mpontillo about this on IRC, and we tried a few things with no success. I booted a failed node with init=/bin/bash and discovered that /etc/netplan was empty rather than containing 50-cloud-init.yaml as expected.
I've run cloud-init collect-logs and updated the result (Canonical-only link): https:/
I think the issue is quite simple actually.
Bionic uses net plan, everything before that uses e/n/i. In trusty, curtin wrote the network configuration. In latest versions, Curtin now doesn’t do the network configuration, and actually tells cloud-unit to do it by passing the configuration.
So what I think is happening is that curtin writes e/n/i and cloud-unit actually never writes net plan.
So I would say that bionic deployment is not supported on 1.9 due to curtin.