search

clemenko / rke_airgap_install

a script/method for air gapping the Rancher Stack with Hauler
50 stars 25 forks source link

Image missing for the deployment of rancher after updating version #10

Closed MozeBaltyk closed 9 months ago

MozeBaltyk commented 10 months ago

Hello Clemenko

Since update on rancher version to 2.8.0, it need some mirrored image:

manager  ●  localhost:5000/rancher/mirrored-cluster-api-controller:v1.4.4  false  ImagePullBackOff

which is missing in the zst package due to exclusion from the list: 

# remove things that are not needed and overlapped
  sed -i -E '/neuvector|minio|gke|aks|eks|sriov|harvester|mirrored|longhorn|thanos|tekton|istio|hyper|jenkins|windows/d' rancher/orig_rancher-images.txt

for this one :

 patch-sa-cattle-provisioning-capi-system  ●  localhost:5000/rancher/kubectl:v1.20.2  false  ImagePullBackOff

I think is excluded with the sorting and keeping only the latest

regards,

clemenko commented 10 months ago

There currently is an issue with upstream v2.8.0. Once it is fixed I will retest and fix the script.

clemenko commented 10 months ago

v2.8.1 came out. I am going to test today.

clemenko commented 10 months ago

Take a look at line 95 and 96.

MozeBaltyk commented 10 months ago

I did something similar in my code:

- name: Rancher List - Due to version 2.8.0
  ansible.builtin.shell:
    cmd: |
      set -o pipefail
      echo "rancher/mirrored-cluster-api-controller:v1.4.4" >> {{ directory_package }}/images/rancher/images.txt
      echo "rancher/kubectl:v1.20.2" >> {{ directory_package }}/images/rancher/images.txt
    executable: /bin/bash
  changed_when: false

Now I am working with "renovate" to autoupdate rke2 and all dependencies. Then, I will work on CI and test.

I did not had time to test the new rancher version but seems that 2.8.0 is to avoid and new to jump immediatly to 2.8.1

So did you had opportunity to test the RKub collection ?

By the way, where did you got the registry image ? is it a custom image or you got it from some registry ? I would like to put it in my renovate.json as well ...

clemenko commented 10 months ago

that is the upstream registry. I was pulling it then updating. I just fixed/updated the code to pull directly from docker hub.

dmeiselman commented 9 months ago

tested yesterday on Rocky 8.9, still seeing helm deployment get stuck on capi-controller-manager failing to pull mirrored-cluster-api-controller:v1.4.4 image.

Oh also: you're still telling workers to install selinux-0.16 not 0.17. Might be worth declaring a $SELINUX_VERSION

clemenko commented 9 months ago

I'll take a look in the morning.

dmeiselman commented 9 months ago

we add cluster-api-controller to the list of images in cleaned_orig_rancher-images.txt, but not into rancher-images.txt. The skopeo operation at line 126 uses rancher-images.txt

clemenko commented 9 months ago

got it sorted. Also for el8! Check the latest.

clemenko commented 9 months ago

confirmed working with several people. feel free to reopen if it is still an issue.