stacks-instances/otc/observability.buildth.ing/stacks/observability/grafana-operator/manifests/cronjob-monitoring.yaml

apiVersion: grafana.integreatly.org/v1beta1
kind: GrafanaDashboard
metadata:
  name: cronjob-monitoring
spec:
  instanceSelector:
    matchLabels:
      dashboards: "grafana"
  json: |
    {
      "annotations": {"list": []},
      "editable": true,
      "graphTooltip": 1,
      "panels": [
        {
          "collapsed": false,
          "gridPos": {"h": 1, "w": 24, "x": 0, "y": 0},
          "title": "Backup Job Status",
          "type": "row"
        },
        {
          "datasource": {"type": "prometheus"},
          "fieldConfig": {"defaults": {"unit": "s", "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "yellow", "value": 86400}, {"color": "red", "value": 172800}]}}},
          "gridPos": {"h": 5, "w": 12, "x": 0, "y": 1},
          "title": "Time Since Last Schedule",
          "type": "stat",
          "targets": [{"expr": "time() - kube_cronjob_status_last_schedule_time{cluster_environment=~\"$cluster_environment\"}", "legendFormat": "{{cronjob}} ({{cluster_environment}})"}]
        },
        {
          "datasource": {"type": "prometheus"},
          "fieldConfig": {"defaults": {"unit": "short", "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 1}]}}},
          "gridPos": {"h": 5, "w": 12, "x": 12, "y": 1},
          "title": "Failed Jobs (Active)",
          "type": "stat",
          "targets": [{"expr": "sum by(cluster_environment, job_name) (kube_job_status_failed{cluster_environment=~\"$cluster_environment\"}) > 0", "legendFormat": "{{job_name}} ({{cluster_environment}})"}]
        },
        {
          "collapsed": false,
          "gridPos": {"h": 1, "w": 24, "x": 0, "y": 6},
          "title": "CronJob Overview",
          "type": "row"
        },
        {
          "datasource": {"type": "prometheus"},
          "fieldConfig": {"defaults": {"custom": {"filterable": true}}, "overrides": [{"matcher": {"id": "byName", "options": "Suspended"}, "properties": [{"id": "mappings", "value": [{"options": {"0": {"text": "No", "color": "green"}, "1": {"text": "YES", "color": "red"}}, "type": "value"}]}]}]},
          "gridPos": {"h": 8, "w": 24, "x": 0, "y": 7},
          "title": "All CronJobs",
          "type": "table",
          "targets": [
            {"expr": "kube_cronjob_info{cluster_environment=~\"$cluster_environment\"}", "format": "table", "instant": true, "refId": "A"}
          ],
          "transformations": [
            {"id": "filterFieldsByName", "options": {"include": {"names": ["cluster_environment", "cronjob", "namespace", "schedule"]}}},
            {"id": "organize", "options": {"renameByName": {"cluster_environment": "Environment", "cronjob": "CronJob", "namespace": "Namespace", "schedule": "Schedule"}}}
          ]
        },
        {
          "collapsed": false,
          "gridPos": {"h": 1, "w": 24, "x": 0, "y": 15},
          "title": "Job History",
          "type": "row"
        },
        {
          "datasource": {"type": "prometheus"},
          "fieldConfig": {"defaults": {"unit": "short"}},
          "gridPos": {"h": 8, "w": 12, "x": 0, "y": 16},
          "title": "Job Completions (24h)",
          "type": "timeseries",
          "targets": [{"expr": "sum(kube_job_status_succeeded{cluster_environment=~\"$cluster_environment\"}) by (job_name, cluster_environment)", "legendFormat": "{{job_name}} ({{cluster_environment}})"}]
        },
        {
          "datasource": {"type": "prometheus"},
          "fieldConfig": {"defaults": {"unit": "short", "color": {"mode": "palette-classic"}}},
          "gridPos": {"h": 8, "w": 12, "x": 12, "y": 16},
          "title": "Job Failures (24h)",
          "type": "timeseries",
          "targets": [{"expr": "sum(kube_job_status_failed{cluster_environment=~\"$cluster_environment\"}) by (job_name, cluster_environment)", "legendFormat": "{{job_name}} ({{cluster_environment}})"}]
        }
      ],
      "schemaVersion": 39,
      "tags": ["edp", "backup", "cronjob"],
      "templating": {
        "list": [
          {
            "current": {"selected": true, "text": "All", "value": "$__all"},
            "datasource": {"type": "prometheus"},
            "definition": "label_values(kube_cronjob_info, cluster_environment)",
            "includeAll": true,
            "multi": true,
            "name": "cluster_environment",
            "label": "Environment",
            "query": "label_values(kube_cronjob_info, cluster_environment)",
            "refresh": 2,
            "sort": 1,
            "type": "query"
          }
        ]
      },
      "time": {"from": "now-24h", "to": "now"},
      "title": "CronJob & Backup Monitoring",
      "uid": "edp-cronjobs"
    }