Since this new functionality https://issues.redhat.com/browse/MCO-706, when using the assisted-installer to install OCP, the firstboot service should detect that, after some previous work before installation, the  installation kernel arguments are the same as the boot kernel arguments and the installation image is the same as the boot image, hence, skipping the first reboot.

The first reboot is being properly skipped in installations using amd64 images, but when we try to install a cluster using the agent installer and arm64 images the first boot is not skipped because MCO detects (correctly) that the installation image and the boot image are not equal.


After talking with the assisted-installer team, we could see that in their test environment the first reboot is skipped (in the machine-config-daemon-firstboot.service logs), that is the reason why we link this jira task to the agent-installer team, because previously we have discarded MCO and the assisted-installer teams.


 

4.15
 

Always
 

1. Install a 4.15 cluster using this prow job 

 periodic-ci-openshift-openshift-tests-private-release-4.15-arm64-nightly-baremetal-sno-agent-ipv4-static-connected-f7

In the machine-config-daemon-firstboot.service logs we can see that the reboot is not skipped because the installation image is not the same as the boot image.

Nov 20 10:07:01.196413 master-00 podman[4836]: I1120 10:07:01.196355    4895 rpm-ostree.go:263] Linking ostree authfile to /var/lib/kubelet/config.json
Nov 20 10:07:01.196538 master-00 podman[4836]: I1120 10:07:01.196407    4895 rpm-ostree.go:243] Executing rebase to quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:679bf7f555730945f86a363f8fa0929dd0b29aa7269798a46987c7a20679ac9a
Nov 20 10:07:01.196538 master-00 podman[4836]: I1120 10:07:01.196441    4895 update.go:2120] Running: rpm-ostree rebase --experimental ostree-unverified-registry:quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:679bf7f555730945f86a363f8fa0929dd0b29aa7269798a46987c7a20679ac9a
Nov 20 10:07:01.250151 master-00 podman[4836]: Pulling manifest: ostree-unverified-image:docker://quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:679bf7f555730945f86a363f8fa0929dd0b29aa7269798a46987c7a20679ac9a
Nov 20 10:07:02.304101 master-00 podman[4836]: Importing: ostree-unverified-image:docker://quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:679bf7f555730945f86a363f8fa0929dd0b29aa7269798a46987c7a20679ac9a (digest: sha256:679bf7f555730945f86a363f8fa0929dd0b29aa7269798a46987c7a20679ac9a)
Nov 20 10:07:02.304101 master-00 podman[4836]: ostree chunk layers stored: 23 needed: 28 (524.4?MB)
Nov 20 10:07:02.304695 master-00 podman[4836]: Fetching ostree chunk sha256:958cecd86de3 (104.4?MB)
Nov 20 10:07:05.125152 master-00 podman[4836]: Fetched ostree chunk sha256:958cecd86de3
Nov 20 10:07:05.125152 master-00 podman[4836]: Fetching ostree chunk sha256:aba65a40a3ae (59.6?MB)
Nov 20 10:07:06.485002 master-00 podman[4836]: Fetched ostree chunk sha256:aba65a40a3ae
Nov 20 10:07:06.485002 master-00 podman[4836]: Fetching ostree chunk sha256:a47347806c81 (29.9?MB)
Nov 20 10:07:07.913075 master-00 podman[4836]: Fetched ostree chunk sha256:a47347806c81
Nov 20 10:07:07.913075 master-00 podman[4836]: Fetching ostree chunk sha256:aeb622bdead9 (17.2?MB)

....

Nov 20 10:07:32.655286 master-00 podman[4836]: Changes queued for next boot. Run "systemctl reboot" to start a reboot
...

 

The installation image should be the same as the boot image and MCO should be able to safely skip the first reboot.
 

In the first comment you can find a tar file including:
- the journal log
- the must-gather file
- the assisted installer service logs (the logs that we could get, since it is not easy to get them because the service is shutdown when it finishes)