On cluster building stuck in pending state

[drriguz]

Hi,
I'm trying to build a knative function with --remote, however the pipeline stucks in pending state, what should I do to fix it?

$ func deploy --remote --service-account agent-less-builder                              
Warning: namespace chosen is 'agent-less', but currently active namespace is 'default'. Continuing with deployment to 'agent-less'.
Creating Pipeline resourcesRunning Pipeline Task: Fetching git repository with the function source code

The pipeline and taskun pods are created successfully, and it says that "0/1 nodes are available: 1 node(s) didn''t match pod affinity rules.".

conditions:
  - lastTransitionTime: '2024-03-27T09:17:59Z'
    message: 'pod status "PodScheduled":"False"; message: "0/1 nodes are available: 1 node(s) didn''t match pod affinity rules. preemption: 0/1 nodes are available: 1 Preemption is not helpful for scheduling.."'
    reason: Pending
    status: Unknown
    type: Succeeded
podName: bing-search-tool-pack-git-pipeline-run-9x49m-fetch-sources-pod

After digging into it, I find that the issue might be caused by the workspaces that used in the pipelinerun:

apiVersion: tekton.dev/v1
kind: PipelineRun
metadata:
  annotations:
    manifestival: new
  generateName: helloworld-pack-git-pipeline-run-
  labels:
    boson.dev/function: 'true'
    boson.dev/runtime: node
    function.knative.dev: 'true'
    function.knative.dev/name: helloworld
    function.knative.dev/runtime: node
  namespace: default
spec:
  params:
    - name: gitRepository
      value: https://github.com/drriguz/helloworld.git
    - name: gitRevision
      value: master
    - name: contextDir
      value: ''
    - name: imageName
      value: localhost:5000/helloworld/helloworld:latest
    - name: registry
      value: localhost:5000/helloworld
    - name: builderImage
      value: ghcr.io/knative/builder-jammy-base:latest
    - name: buildEnvs
      value:
        - '='
  pipelineRef:
    name: helloworld-pack-git-pipeline
  taskRunTemplate:
    serviceAccountName: default
  timeouts:
    pipeline: 1h0m0s
  workspaces:
    - name: source-workspace
      persistentVolumeClaim:
        claimName: helloworld-pack-git-pipeline-pvc
      subPath: source
    - name: cache-workspace
      persistentVolumeClaim:
        claimName: helloworld-pack-git-pipeline-pvc
      subPath: cache
    - name: dockerconfig-workspace
      secret:
        secretName: helloworld-pack-git-pipeline-secret

However, because I'm running it using minikube, pvc-04c96a5c-2916-49b5-9909-34b7b7998838 is use hostpath storageclass and it's RWO, however it's shared by both source-workspace and cache-workspace, which then cause the pod getting stuck waiting for the PVC to become available.

$ kubectl get pvc
helloworld-pack-git-pipeline-pvc   Bound    pvc-04c96a5c-2916-49b5-9909-34b7b7998838   256Mi      RWO            standard

[drriguz]

It seems that this could be solved by disable affinity assistant:

kubectl edit configmap feature-flags -n tekton-pipelines
# set disable-affinity-assistant to true

Ref: https://tekton.dev/docs/pipelines/affinityassistants/