`taskrunCancelled` sets the `startedAt` and `finishedAt` times for steps that were not executed

[pritidesai]

Expected Behavior

When a taskRun is cancelled due to a timeout, I expect the steps that were not executed to have no timings attached to them. It is misleading to set the same time for all steps, as steps are generally executed in sequence. Additionally, the startedAt time is set to the start time of the taskRun. Meaning, if a taskRun is canceled while executing the second step of a task, that step may have a startedAt time earlier than the first step, which is inaccurate.

Actual Behavior

Steps to Reproduce the Problem

apiVersion: tekton.dev/v1beta1
kind: Task
metadata:
  name: task
spec:
  steps:
    - image: mirror.gcr.io/busybox
      name: step-1
      command: ['/bin/sh']
      args: ['-c', 'sleep 1']
    - image: mirror.gcr.io/busybox
      name: step-2
      command: ['/bin/sh']
      args: ['-c', 'sleep 10']
    - image: mirror.gcr.io/busybox
      name: step-3
      command: ['/bin/sh']
      args: ['-c', 'sleep 10']
    - image: mirror.gcr.io/busybox
      name: step-4
      command: ['/bin/sh']
      args: ['-c', 'sleep 10']
---

apiVersion: tekton.dev/v1beta1
kind: Pipeline
metadata:
  name: pipeline
spec:
  tasks:
    - name: foo
      taskRef:
        name: task
---

apiVersion: tekton.dev/v1beta1
kind: PipelineRun
metadata:
  name: pipelinerun
spec:
  pipelineRef:
    name: pipeline
  timeouts:
    pipeline: 10s

Apply this pipelineRun:

k create -f pipelinerun-timeout.yaml

Analyze the taskRun:

 tkn taskrun describe pipelinerun-foo                    
Name:              pipelinerun-foo
Namespace:         default
Task Ref:          task
Service Account:   default
Timeout:           24h0m0s
Labels:
 app.kubernetes.io/managed-by=tekton-pipelines
 tekton.dev/memberOf=tasks
 tekton.dev/pipeline=pipeline
 tekton.dev/pipelineRun=pipelinerun
 tekton.dev/pipelineTask=foo
 tekton.dev/task=task
Annotations:
 pipeline.tekton.dev/release=e58ca13

🌡️  Status

STARTED          DURATION    STATUS
11 seconds ago   10s         Cancelled(TaskRunCancelled)

Message

TaskRun "pipelinerun-foo" was cancelled. TaskRun cancelled as the PipelineRun it belongs to has timed out.

🦶 Steps

 NAME       STATUS
 ∙ step-2   TaskRunCancelled
 ∙ step-3   TaskRunCancelled
 ∙ step-4   TaskRunCancelled
 ∙ step-1   Completed

Analyze the steps:

k get taskruns.tekton.dev pipelinerun-foo -o json | jq .status.steps
[
  {
    "container": "step-step-1",
    "imageID": "mirror.gcr.io/busybox@sha256:2919d0172f7524b2d8df9e50066a682669e6d170ac0f6a49676d54358fe970b5",
    "name": "step-1",
    "terminated": {
      "containerID": "containerd://0f9896d9080b86b7deb09dea2b2b73d3f5d5a8a57a5e76222df9b3dce911ff1b",
      "exitCode": 0,
      "finishedAt": "2025-01-10T20:07:18Z",
      "reason": "Completed",
      "startedAt": "2025-01-10T20:07:17Z"
    },
    "terminationReason": "Completed"
  },
  {
    "container": "step-step-2",
    "imageID": "mirror.gcr.io/busybox@sha256:2919d0172f7524b2d8df9e50066a682669e6d170ac0f6a49676d54358fe970b5",
    "name": "step-2",
    "terminated": {
      "exitCode": 1,
      "finishedAt": "2025-01-10T20:07:22Z",
      "message": "Step step-2 terminated as pod pipelinerun-foo-pod is terminated",
      "reason": "TaskRunCancelled",
      "startedAt": "2025-01-10T20:07:15Z"
    },
    "terminationReason": "TaskRunCancelled"
  },
  {
    "container": "step-step-3",
    "imageID": "mirror.gcr.io/busybox@sha256:2919d0172f7524b2d8df9e50066a682669e6d170ac0f6a49676d54358fe970b5",
    "name": "step-3",
    "terminated": {
      "exitCode": 1,
      "finishedAt": "2025-01-10T20:07:22Z",
      "message": "Step step-3 terminated as pod pipelinerun-foo-pod is terminated",
      "reason": "TaskRunCancelled",
      "startedAt": "2025-01-10T20:07:15Z"
    },
    "terminationReason": "TaskRunCancelled"
  },
  {
    "container": "step-step-4",
    "imageID": "mirror.gcr.io/busybox@sha256:2919d0172f7524b2d8df9e50066a682669e6d170ac0f6a49676d54358fe970b5",
    "name": "step-4",
    "terminated": {
      "exitCode": 1,
      "finishedAt": "2025-01-10T20:07:22Z",
      "message": "Step step-4 terminated as pod pipelinerun-foo-pod is terminated",
      "reason": "TaskRunCancelled",
      "startedAt": "2025-01-10T20:07:16Z"
    },
    "terminationReason": "TaskRunCancelled"
  }
]

Step	startedAt	finishedAt
step-1	2025-01-10T20:07:17Z	2025-01-10T20:07:18Z
step-2	2025-01-10T20:07:15Z	2025-01-10T20:07:22Z
step-3	2025-01-10T20:07:15Z	2025-01-10T20:07:22Z
step-4	2025-01-10T20:07:16Z	2025-01-10T20:07:22Z

Additional Info

• Kubernetes version:

  Output of kubectl version:

 k version
Client Version: v1.31.0
Server Version: v1.29.11+IKS

• Tekton Pipeline version:

  Output of tkn version or kubectl get pods -n tekton-pipelines -l app=tekton-pipelines-controller -o=jsonpath='{.items[0].metadata.labels.version}'

 tkn version                                                                                                                               
Client version: 0.32.0
Pipeline version: v0.62.1

[vdemeester]

Marking as good first issue — this is a focused bug fix in one file.

What's happening

When a TaskRun is cancelled (e.g., due to timeout), all steps — including ones that never ran — get startedAt and finishedAt timestamps. Steps that never executed should have no timing information.

Where to look

• Main logic: pkg/pod/status.go — the MakeTaskRunStatus function and the step state construction loop (around line 380-400)
• When building StepState objects, the code copies container state including timestamps from containers that were never actually started
• The fix should check whether a step's container was actually started before copying timing information

What the fix should do

• If a step container was never started (i.e., its state is only Terminated because the pod was killed, not because the step ran), don't set StartedAt / FinishedAt
• Or: set them to zero values for steps that have no Running or Terminated state with a real start time

How to verify

• Add a unit test in pkg/pod/status_test.go with a cancelled TaskRun where only the first step ran
• Verify that unexecuted steps have zero-value timestamps
• Run: go test ./pkg/pod/...

/good-first-issue

[adityavshinde]

@waveywaves - I would like to work on this issue.

[adityavshinde]

/assign @adityavshinde