Kubernetes CronJob and Batch Job Execution Failure Sentinel

id: kubernetes-cronjob-execution-failure-sentinel
namespace: company.sre

inputs:
  - id: kube_context
    type: STRING
    required: false
    description: Optional Kubernetes context name.
  - id: max_failed_cronjobs
    type: INT
    defaults: 0
    description: Maximum acceptable count of CronJobs exhibiting missing execution
      or failed active pods.

tasks:
  - id: audit_cronjobs
    type: io.kestra.plugin.scripts.shell.Commands
    containerImage: bitnami/kubectl:latest
    commands:
      - |
        CONTEXT_ARG=""
        if [ -n "{{ inputs.kube_context }}" ]; then
          CONTEXT_ARG="--context={{ inputs.kube_context }}"
        fi

        kubectl $CONTEXT_ARG get cronjobs -A -o json > /tmp/cronjobs_all.json
        kubectl $CONTEXT_ARG get jobs -A -o json > /tmp/jobs_all.json

        FAILED_JOBS=$(jq '[.items[] | select(.status.failed != null and .status.failed > 0) | {namespace: .metadata.namespace, job_name: .metadata.name, failed: .status.failed}]' /tmp/jobs_all.json)
        FAILED_COUNT=$(echo "$FAILED_JOBS" | jq 'length')

        echo "::set-output name=failed_count::$FAILED_COUNT::"
        echo "::set-output name=failed_jobs::$FAILED_JOBS::"

  - id: evaluate_cronjob_health
    type: io.kestra.plugin.core.flow.If
    condition: "{{ outputs.audit_cronjobs.vars.failed_count | int >
      inputs.max_failed_cronjobs }}"
    then:
      - id: notify_operations_team
        type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
        url: "{{ secret('SLACK_WEBHOOK') }}"
        payload: |
          {
            "channel": "#infrastructure-alerts",
            "attachments": [
              {
                "color": "danger",
                "title": "🚨 Kubernetes Scheduled Job / CronJob Failure Alert",
                "text": "Detected *{{ outputs.audit_cronjobs.vars.failed_count }}* failed batch Job execution(s) across cluster namespaces. Failed jobs: {{ outputs.audit_cronjobs.vars.failed_jobs }}. Impact: Scheduled database backups, ETL reconciliations, or maintenance cleanups did not execute successfully.",
                "footer": "Kestra SRE Sentinel • CronJob Execution"
              }
            ]
          }

errors:
  - id: alert_pipeline_failure
    type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
    url: "{{ secret('SLACK_WEBHOOK') }}"
    payload: |
      {
        "channel": "#infrastructure-alerts",
        "attachments": [
          {
            "color": "danger",
            "title": "⚠️ CronJob Sentinel Execution Failed",
            "text": "The Kubernetes CronJob audit workflow encountered an execution error. Please verify cluster RBAC permissions and kubeconfig.",
            "footer": "Kestra Workflow Failure Sentinel"
          }
        ]
      }

triggers:
  - id: periodic_schedule
    type: io.kestra.plugin.core.trigger.Schedule
    cron: "0 * * * *"
    disabled: true