OS update is never triggered

[ldevulder]

What steps did you take and what happened:
Create a new OS channel pointing to newer image versions, create an update group to upgrade the OS on the nodes.

What did you expect to happen:
OS image on the nodes updated from SLEMicro 6.0 to 6.1.

Anything else you would like to add:
Find elemental-operator logs: upgrade.log.

Environment:

• Elemental release version (use cat /etc/os-release):

NAME="SL-Micro"
VERSION="6.0"
VERSION_ID="6.0"
PRETTY_NAME="SUSE Linux Micro 6.0"
ID="sl-micro"
ID_LIKE="suse"
ANSI_COLOR="0;32"
CPE_NAME="cpe:/o:suse:sl-micro:6.0"
HOME_URL="https://www.suse.com/products/micro/"
DOCUMENTATION_URL="https://documentation.suse.com/sl-micro/6.0/"
IMAGE_REPO="registry.suse.com/suse/sl-micro/6.0/baremetal-os-container"
IMAGE_TAG="2.1.3-4.7"
IMAGE="registry.suse.com/suse/sl-micro/6.0/baremetal-os-container:2.1.3-4.7"
TIMESTAMP=20241107104849
GRUB_ENTRY_NAME="SUSE Linux Micro"

• Rancher version: Prime v2.10.2
• Kubernetes version (use kubectl version): v1.30.6+k3s1 for Rancher Manager and v1.30.8+k3s1 for the Elemental cluster
• Cloud provider or hardware configuration: Virtual Machines.

How to reproduce the issue:

• Deploy Rancher Manager Stable or Prime version
• Create a simple Elemental cluster (1 node is enough) with Stable version (operator v1.6.5, OS channel v6.0)
• Add the Staging OS channel:

apiVersion: elemental.cattle.io/v1beta1
kind: ManagedOSVersionChannel
metadata:
  name: sl-micro-6.1-baremetal-channel
  namespace: fleet-default
spec:
  options:
    image: registry.opensuse.org/isv/rancher/elemental/staging/containers/rancher/elemental-channel/sl-micro:6.1-baremetal
  syncInterval: 1h
  type: custom

• Create an update group:

apiVersion: elemental.cattle.io/v1beta1
kind: ManagedOSImage
metadata:
  name: upgrade-os-to-6.1
  namespace: fleet-default
spec:
  clusterTargets:
    - clusterName: test-cluster
  managedOSVersionName: baremetal-v2.2.0-3.7-os

• All looks good but the update is never triggered

[ldevulder]

I tested with osImage option instead of managedOSVersionName and it worked BUT ONLY with dev instead of staging:

apiVersion: elemental.cattle.io/v1beta1
kind: ManagedOSImage
metadata:
  name: upgrade-os
  namespace: fleet-default
spec:
  clusterTargets:
    - clusterName: test-cluster
  osImage: registry.opensuse.org/isv/rancher/elemental/dev/containers/suse/sl-micro/6.1/baremetal-os-container:latest

I checked and staging is available on registry.opensuse.org, like dev.

[ldevulder]

I retested today and it worked with update-os-to-6-1 as the update group name but it never start with update-to-os-6.1, so clearly it's not a name length issue but because of the point in the naming.

Looks like more a Rancher Manager issue, no?

[fgiudici]

The issue is with the dot . in the resource name.
This tracks down to the system-upgrade-controller resources created under the hood.

Tried to create a ManagedOSImage resource with name test.name.

Checking the system-upgrade-controller logs on the child cluster:

kubectl -n cattle-system logs deployment/system-upgrade-controller

time="2025-02-18T08:42:50Z" level=error msg="error syncing 'cattle-system/os-upgrader-test.name': handler system-upgrade-controller: failed to create cattle-system/apply-os-upgrader-test.name-on-m-9f544b7d-e272-4fe5-9db2-2494f8 batch/v1, Kind=Job for system-upgrade-controller cattle-system/os-upgrader-test.name: Job.batch \"apply-os-upgrader-test.name-on-m-9f544b7d-e272-4fe5-9db2-2494f8\" is invalid: [spec.template.spec.volumes[2].name: Invalid value: \"secret-os-upgrader-test.name\": must not contain dots, spec.template.spec.containers[0].volumeMounts[2].name: Not found: \"secret-os-upgrader-test.name\", spec.template.spec.initContainers[0].volumeMounts[2].name: Not found: \"secret-os-upgrader-test.name\"], requeuing"

So, guess we should not allow dots . in the upgrade groups (ManagedOSImage resources).
Or at the very minimum document this.