Regression with PVCs and node affinity

[jpds]

Expected Behavior

I have a pipeline that works fine in 0.12.1 but fails on 0.13.0.

Actual Behavior

Pod for pipeline fails with:

Events:
  Type     Reason             Age                From                Message
  ----     ------             ----               ----                -------
  Normal   NotTriggerScaleUp  72s                cluster-autoscaler  pod didn't trigger scale-up (it wouldn't fit if a new node is added): 1 node(s) didn't match pod affinity/anti-affinity, 1 node(s) didn't match pod affinity rules, 1 node(s) didn't match node selector
  Warning  FailedScheduling   70s (x4 over 75s)  default-scheduler   0/1 nodes are available: 1 node(s) didn't match node selector.

Steps to Reproduce the Problem

1. Deploy 0.13.0:

kubectl apply -f https://storage.googleapis.com/tekton-releases/pipeline/previous/v0.13.0/release.yaml

2. Apply this pipeline:

apiVersion: tekton.dev/v1beta1
kind: Task
metadata:
  name: test-task
spec:
  steps:
    - name: test
      image: alpine:3.11.6
      script: |
        #!/bin/sh
        set -e
        echo "$(date -R): Test"
  workspaces:
  - name: test-dir
---
apiVersion: tekton.dev/v1beta1
kind: Pipeline
metadata:
  name: volume-pipeline
spec:
  tasks:
    - name: test
      taskRef:
        name: test-task
      workspaces:
        - name: test-dir
          workspace: volume-workspace
  workspaces:
  - name: volume-workspace
---
apiVersion: tekton.dev/v1beta1
kind: PipelineRun
metadata:
  name: volume-pipelinerun
spec:
  pipelineRef:
    name: volume-pipeline
  podTemplate:
    nodeSelector:
      cloud.google.com/gke-nodepool: test-pool
    volumes:
    - name: volume-storage
  timeout: 2h
  workspaces:
    - name: volume-workspace
      volumeClaimTemplate:
        metadata:
          name: mypvc
        spec:
          accessModes:
            - ReadWriteOnce
          resources:
            requests:
              storage: 2Gi

3. Pod fails to start with above error.

4. Remove 0.13.0.

5. Install 0.12.1

6. Apply above pipeline.

7. Autoscaling succeeds:

Events:
  Type     Reason            Age                From                Message
  ----     ------            ----               ----                -------
  Warning  FailedScheduling  11s (x3 over 15s)  default-scheduler   0/1 nodes are available: 1 node(s) didn't match node selector.
  Normal   TriggeredScaleUp  11s                cluster-autoscaler  pod triggered scale-up: [{https://content.googleapis.com/compute/v1/projects/project-jklasdfk334/zones/us-west1-a/instanceGroups/gke-test-regression-test-pool-48f4fc0a-grp 0->1 (max: 1)}]

Additional Info

• Kubernetes version: 1.16.8-gke.15

• Tekton Pipeline version: 0.13.0

[jlpettersson]

1 node(s) didn't match pod affinity/anti-affinity, 1 node(s) didn't match pod affinity rules

Hi, yes, this is a known bug.

Unfortunately the name of PipelineRun + Workspace, can only be up to 34 chars long. In your case, it is 35 chars. You can use shorter PipelineRun Name or Workspace Name, or disable the affinity assistant

The bug will be fixed by #2768 in 0.13.1

This is a duplicate of #2766

[vdemeester]

/kind bug

[bobcatfish]

Closing as duplicate of #2766, please re-open if there is a new issue here that we missed

[aeweidne]

I am experiencing this on 0.14.0 -- nevermind, hitting this one #2829