#Disable Live Collection of Incomplete Runs

#Feature Overview

By default, the tekton-results-watcher stores TaskRun and PipelineRun records when the runs are created, and then continuously upserts those records whenever the resources change. This gives live visibility into running resources, but it also increases write load on the Results API and database.

You can reduce this write load by setting the --disable_storing_incomplete_runs=true flag on the tekton-results-watcher controller. When this flag is enabled, incomplete runs are not stored in Tekton Results. The watcher stores a run only after it reaches a completed state.

#Use Cases

• Reduce Results database writes in clusters with many short-lived or frequently updated runs.
• Improve watcher and database performance when live visibility for running resources is not required.
• Store only final execution records while keeping completed run history available through Tekton Results.

#Prerequisites

• Tekton Results is installed through TektonConfig.
• You have permission to update the TektonConfig resource.
• You know the namespace where Results is installed. The examples below use the default tekton-pipelines namespace.

#Steps

#1. Configure the watcher flag

Update TektonConfig.spec.result.options.deployments.tekton-results-watcher and add the flag to the watcher container:

apiVersion: operator.tekton.dev/v1alpha1
kind: TektonConfig
metadata:
  name: config
spec:
  result:
    options:
      deployments:
        tekton-results-watcher:
          spec:
            template:
              spec:
                containers:
                  - name: watcher
                    args:
                      - "--disable_storing_incomplete_runs=true"

The operator merges container arguments by flag key. You do not need to copy the existing watcher arguments such as -api_addr, -auth_mode, or -logs_api; the new flag is added to the existing argument list.

#2. Apply the configuration

Apply the updated TektonConfig manifest:

$ kubectl apply -f tekton-config-disable-incomplete-runs.yaml

tektonconfig.operator.tekton.dev/config configured

Wait for the watcher Deployment to roll out:

$ kubectl rollout status -n tekton-pipelines deployment/tekton-results-watcher

deployment "tekton-results-watcher" successfully rolled out

#3. Verify the watcher arguments

Check that the watcher pod template includes the flag:

$ kubectl get deployment -n tekton-pipelines tekton-results-watcher \
  -o jsonpath='{range .spec.template.spec.containers[?(@.name=="watcher")].args[*]}{.}{"\n"}{end}'

-api_addr
$(TEKTON_RESULTS_API_SERVICE)
-auth_mode
$(AUTH_MODE)
-logs_api=true
--disable_storing_incomplete_runs=true

#Operation Results

After the rollout finishes:

• Running TaskRun and PipelineRun resources are not stored in Tekton Results while their status is still incomplete.
• Completed TaskRun and PipelineRun resources are stored after they finish.
• The Results API and database receive fewer updates because intermediate resource status changes are not continuously upserted.

#Functional Verification

Create or trigger a TaskRun or PipelineRun that takes long enough to inspect while it is running.

While the run is still incomplete, query Results records:

$ tkn results records list default/results/-

The running resource should not appear in the Results records.

After the run completes, query the records again:

$ tkn results records list default/results/-

The completed resource should now appear in the Results records.

#Roll Back

To restore live collection, set the flag to false:

apiVersion: operator.tekton.dev/v1alpha1
kind: TektonConfig
metadata:
  name: config
spec:
  result:
    options:
      deployments:
        tekton-results-watcher:
          spec:
            template:
              spec:
                containers:
                  - name: watcher
                    args:
                      - "--disable_storing_incomplete_runs=false"

Apply the updated TektonConfig and wait for the watcher Deployment to roll out again:

$ kubectl apply -f tekton-config-enable-live-collection.yaml
$ kubectl rollout status -n tekton-pipelines deployment/tekton-results-watcher

#Troubleshooting

#The flag is not visible in the watcher Deployment

Check whether the TektonConfig update was accepted:

$ kubectl get tektonconfig config -o yaml

Confirm that the configuration is under spec.result.options.deployments.tekton-results-watcher and that the container name is watcher.

#Running resources still appear in Results

Check whether the new watcher pod is running:

$ kubectl get pods -n tekton-pipelines -l app.kubernetes.io/name=tekton-results-watcher

If an older pod is still running, wait for the rollout to finish. Also verify that the run you are checking was created after the new watcher pod started.

#Completed resources do not appear in Results

Check the watcher logs for storage or API errors:

$ kubectl logs -n tekton-pipelines deployment/tekton-results-watcher

Also confirm that the Results API and database are healthy. For installation and basic verification steps, see Quick Start.