#Regular Cleanup of TaskRun and PipelineRun Resources

#Step 1

Edit the TektonConfig resource

$ kubectl edit tektonconfigs.operator.tekton.dev config

#Step 2

Modify the spec.pruner configuration, as shown below:

apiVersion: operator.tekton.dev/v1alpha1
kind: TektonConfig
metadata:
  name: config
spec:
  pruner:
    # Whether to disable the pruner functionality, defaults to false
    disabled: false
    # Cron expression for scheduled cleanup
    schedule: "0 8 * * *"
    # Optional: Deadline for task start (in seconds)
    startingDeadlineSeconds: 100
    # Types of resources to clean up
    resources:
      - taskrun
      - pipelinerun
    # Number of resources to retain
    keep: 3
    # Apply cleanup strategies separately for each resource type
    prune-per-resource: true
    # Retain resources within the specified time (in minutes)
    # keep-since: 1440
    # Note: only one of keep and keep-since can be used

Key configuration descriptions:

• disabled: When set to true, disables the pruner functionality (default value: false)
• schedule: Execution schedule for the cleanup task, using cron expression format
• startingDeadlineSeconds: Optional configuration for the deadline for task start (in seconds). If the task misses the scheduled time and exceeds this duration, it is counted as failed
• resources: Supported resource types for cleanup, including taskrun and pipelinerun
• keep: Maximum number of resources to retain during cleanup
• keep-since: Retain resources within a specified time (in minutes)
• prune-per-resource: When set to true (default value is false), applies the keep configuration separately for each resource
  • For example: If there are two Pipelines in namespace ns-1, named pipeline-1 and pipeline-2, the following will be executed:

    tkn pipelinerun delete --pipeline=my-pipeline-1 --keep=3 --namespace=ns-1
    tkn pipelinerun delete --pipeline=my-pipeline-2 --keep=3 --namespace=ns-1
  • Note: When using keep-since, enabling prune-per-resource=true has no practical meaning, as keep-since is based on time constraint for resources, unrelated to the quantity of resources.

#Step 3

Configure cleanup strategies via namespace annotations (optional)

By default, the pruner task will use the global configuration (spec.pruner), but users can customize the pruner configuration for specific namespaces through the following annotations. If certain annotations are missing or have invalid values, global configuration values will be used, or the namespace will be skipped.

Supported annotations:

• operator.tekton.dev/prune.skip: When set to true, skips the pruner task for that namespace
• operator.tekton.dev/prune.schedule: Sets the cleanup schedule for that namespace
• operator.tekton.dev/prune.keep: Sets the maximum number of resources to retain
• operator.tekton.dev/prune.keep-since: Sets the retention period for resources within a specified time (in minutes)
• operator.tekton.dev/prune.prune-per-resource: Sets whether to apply the cleanup strategy separately for each resource
• operator.tekton.dev/prune.resources: Sets the types of resources to clean up, which can be taskrun and/or pipelinerun, with multiple values separated by commas
• operator.tekton.dev/prune.strategy: Sets the cleanup strategy, with optional values of keep or keep-since

Example:

$ kubectl annotate namespace default \
    operator.tekton.dev/prune.schedule="0 1 * * *" \
    operator.tekton.dev/prune.keep="5" \
    operator.tekton.dev/prune.resources="taskrun,pipelinerun"

::: note If no global configuration is set, the following default values will be used:

• resources: pipelinerun
• keep: 100 :::