titan-iac/services/monitoring/dashboards/atlas-testing.json

{
  "uid": "atlas-testing",
  "title": "Atlas Testing",
  "folderUid": "atlas-internal",
  "editable": true,
  "panels": [
    {
      "id": 1,
      "type": "stat",
      "title": "Glue Jobs Stale (>36h)",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 0,
        "y": 0
      },
      "targets": [
        {
          "expr": "(sum((((time() - (kube_cronjob_status_last_successful_time and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"})) > bool 129600) unless on(namespace,cronjob) (kube_cronjob_spec_suspend and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"}) == 1)) + count(((kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"} unless on(namespace,cronjob) kube_cronjob_status_last_successful_time) unless on(namespace,cronjob) (kube_cronjob_spec_suspend and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"}) == 1)))",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "yellow",
                "value": 1
              },
              {
                "color": "orange",
                "value": 2
              },
              {
                "color": "red",
                "value": 3
              }
            ]
          },
          "unit": "none",
          "custom": {
            "displayMode": "auto"
          }
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 2,
      "type": "table",
      "title": "Glue Jobs Missing Success",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 6,
        "y": 0
      },
      "targets": [
        {
          "expr": "((kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"} unless on(namespace,cronjob) kube_cronjob_status_last_successful_time) unless on(namespace,cronjob) (kube_cronjob_spec_suspend and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"}) == 1)",
          "refId": "A",
          "instant": true
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "none",
          "custom": {
            "filterable": true
          }
        },
        "overrides": []
      },
      "options": {
        "showHeader": true,
        "columnFilters": false
      },
      "transformations": [
        {
          "id": "labelsToFields",
          "options": {}
        },
        {
          "id": "sortBy",
          "options": {
            "fields": [
              "Value"
            ],
            "order": "desc"
          }
        }
      ]
    },
    {
      "id": 3,
      "type": "table",
      "title": "Glue Jobs Suspended",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 12,
        "y": 0
      },
      "targets": [
        {
          "expr": "(kube_cronjob_spec_suspend and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"}) == 1",
          "refId": "A",
          "instant": true
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "none",
          "custom": {
            "filterable": true
          }
        },
        "overrides": []
      },
      "options": {
        "showHeader": true,
        "columnFilters": false
      },
      "transformations": [
        {
          "id": "labelsToFields",
          "options": {}
        },
        {
          "id": "sortBy",
          "options": {
            "fields": [
              "Value"
            ],
            "order": "desc"
          }
        }
      ]
    },
    {
      "id": 4,
      "type": "table",
      "title": "Glue Jobs Active Runs",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 6,
        "x": 18,
        "y": 0
      },
      "targets": [
        {
          "expr": "(kube_cronjob_status_active and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"})",
          "refId": "A",
          "instant": true
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "none",
          "custom": {
            "filterable": true
          }
        },
        "overrides": []
      },
      "options": {
        "showHeader": true,
        "columnFilters": false
      },
      "transformations": [
        {
          "id": "labelsToFields",
          "options": {}
        },
        {
          "id": "sortBy",
          "options": {
            "fields": [
              "Value"
            ],
            "order": "desc"
          }
        }
      ]
    },
    {
      "id": 5,
      "type": "table",
      "title": "Glue Jobs Last Success (hours ago)",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 0,
        "y": 4
      },
      "targets": [
        {
          "expr": "((time() - (kube_cronjob_status_last_successful_time and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"}))) / 3600",
          "refId": "A",
          "instant": true
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "h",
          "custom": {
            "filterable": true
          }
        },
        "overrides": []
      },
      "options": {
        "showHeader": true,
        "columnFilters": false
      },
      "transformations": [
        {
          "id": "labelsToFields",
          "options": {}
        },
        {
          "id": "sortBy",
          "options": {
            "fields": [
              "Value"
            ],
            "order": "desc"
          }
        }
      ]
    },
    {
      "id": 6,
      "type": "table",
      "title": "Glue Jobs Last Schedule (hours ago)",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 8,
        "w": 12,
        "x": 12,
        "y": 4
      },
      "targets": [
        {
          "expr": "((time() - (kube_cronjob_status_last_schedule_time and on(namespace,cronjob) kube_cronjob_labels{label_atlas_bstein_dev_glue=\"true\"}))) / 3600",
          "refId": "A",
          "instant": true
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "h",
          "custom": {
            "filterable": true
          }
        },
        "overrides": []
      },
      "options": {
        "showHeader": true,
        "columnFilters": false
      },
      "transformations": [
        {
          "id": "labelsToFields",
          "options": {}
        },
        {
          "id": "sortBy",
          "options": {
            "fields": [
              "Value"
            ],
            "order": "desc"
          }
        }
      ]
    }
  ],
  "time": {
    "from": "now-7d",
    "to": "now"
  },
  "annotations": {
    "list": []
  },
  "schemaVersion": 39,
  "style": "dark",
  "tags": [
    "atlas",
    "testing"
  ]
}