The Result Watcher is a Kubernetes Controller that watches for changes to certain Tekton types and automatically creates/updates their data in the Result API.
The Watcher currently supports the following types:
tekton.dev/v1beta1 TaskRun
tekton.dev/v1beta1 PipelineRun
The Watcher uses Object data to automatically detect and group related Records into the same Result. The following data is checked (listed in order of precedence):
results.tekton.dev/result
annotation. This should correspond to the fullResult.name
identifier (e.g.foo/results/bar
).triggers.tekton.dev/triggers-eventid
label (this is generated from Objects created via Tekton Triggers)- An OwnerReference to a PipelineRun.
If no annotation is detected, the Watcher will automatically generate a new Result name for the Object.
Users and/or integrators can pass arbitrary keys/values to Results by adding special annotations to PipelineRuns and TaskRuns:
results.tekton.dev/resultAnnotations
: a JSON object (string->string) to be stored into theeResult.Annotations
field.results.tekton.dev/recordSummaryAnnotations
: a JSON object (string->string) to be stored into theeResult.Summary.Annotations
field.
Once the Watcher detects those annotations in the observed object, it passes the keys/values to the respective fields of the underlying Result. Those annotations can be used to store relevant metadata (e.g. the Git commit SHA that triggered a PipelineRun) into Results and may be used later to retrieve the objects from the API server. For instance:
apiVersion: tekton.dev/v1beta1
kind: PipelineRun
metadata:
generateName: hello-run-
annotations:
results.tekton.dev/resultAnnotations: |-
{"repo": "tektoncd/results", "commit": "1a6b908"}
results.tekton.dev/recordSummaryAnnotations: |-
{"foo": "bar"}
When the command line flag is completed_run_grace_period
is set to any value other than 0
, resources will be deleted after the specified duration in the flag, calculated from the time of completion. If the value is < 0
, Runs will be deleted immediately after completion or failure.
The flag check_owner
allows additional check before deleting a resource. If set true
, resources with any owner references set will not be deleted. When the flag is false
, owner references will be not be checked before deletion.