Set Up Slack Alerts for Periodic Job Results

We can set up Slack alerts for job results by defining reporter_config in the job definition.

1
2
3
4
5
6

reporter_config:
  slack:
    channel: '#forum'
    job_states_to_report:
    - failure
    report_template: Job {{.Spec.Job}} failed.

For example, by the above snippet, a Slack alert will be sent out to #forum channel when there is a failure of the job. The alert is formatted by report_template.

• The channel has to be in coreos.slack.com.
• The channel has to be public. If it is not then the @prow bot has to be added to it otherwise it won’t be able to properly post messages.
• The state in job_states_to_report has to be a valid Prow job state. See upstream documentation.
• The value of report_template is a Go template and it will be applied to the Prow job instance. The annotations such as {{.Spec.Job}} will be replaced by the job name when the alert is received in Slack. See upstream documentation for more fields of a Prow job. Note that no alerts will be sent out if the template is broken, e.g., cannot be parsed or applied successfully.

The following snippet shows an example of embedding the reporter_config into a job definition:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

agent: kubernetes
cron: 0 */6 * * *
decorate: true
name: periodic-job-name
reporter_config:
  slack:
    channel: '#forum'
    job_states_to_report:
    - failure
    report_template: Job {{.Spec.Job}} failed.
spec: {}  # Valid Kubernetes PodSpec.

Note that we have to modify the job yaml directly even if the job is generated from a ci-operator’s config and regenerating the jobs does not change the existing reporter_config. Moreover, we do not support reporter_config for generated presubmits and postsubmits.