+++ This bug was initially created as a clone of Bug #1840222 +++
--- Additional comment from Colin Walters on 2020-05-28 17:28:30 UTC ---
Current diagnosis is that something in the IPI stack (probably Ironic) is forcibly powering off machines during the first boot.
Our stack is currently not very robust around this. We need to change the MCO's firstboot handling to be more idempotent:
Tested with 4.5.0-0.nightly-2020-06-17-234944
There's no clear reproduction for this as it happens to be flacky anyway, I've spun up a cluster with the above nightly and made sure the changes to the unit are there (no more BindsTo):
17:25:15 [~/Downloads] export KUBECONFIG=cluster-bot-2020-06-18-142422.kubeconfig
17:25:19 [~/Downloads] oc get nodes
NAME STATUS ROLES AGE VERSION
ip-10-0-163-11.us-west-2.compute.internal Ready master 30m v1.18.3+91d0edd
ip-10-0-163-148.us-west-2.compute.internal Ready master 31m v1.18.3+91d0edd
ip-10-0-177-120.us-west-2.compute.internal Ready worker 20m v1.18.3+91d0edd
ip-10-0-185-241.us-west-2.compute.internal Ready worker 20m v1.18.3+91d0edd
ip-10-0-245-33.us-west-2.compute.internal Ready master 30m v1.18.3+91d0edd
ip-10-0-249-229.us-west-2.compute.internal Ready worker 20m v1.18.3+91d0edd
17:25:24 [~/Downloads] oc debug node ip-10-0-163-11.us-west-2.compute.internal
Error from server (NotFound): pods "node" not found
17:25:30 [~/Downloads] oc debug node/ip-10-0-163-11.us-west-2.compute.internal 1 ↵
Starting pod/ip-10-0-163-11us-west-2computeinternal-debug ...
To use host binaries, run `chroot /host`
Pod IP: 10.0.163.11
If you don't see a command prompt, try pressing enter.
sh-4.2# chroot /host bash
[root@ip-10-0-163-11 /]# systemctl cat machine-config-daemon-firstboot.service
Description=Machine Config Daemon Firstboot
# Make sure it runs only on OSTree booted system
# Removal of this file signals firstboot completion
# We only want to run on 4.3 clusters and above; this came from
# Need oneshot to delay kubelet
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.