Merge "Adding apex-pdp metrics to SLAs dashboard."

1 files changed, 553 insertions, 32 deletions

diff --git a/csit/metrics/dashboards/sla-metrics.json b/csit/metrics/dashboards/sla-metrics.json
index 7b96c935..6093b51f 100644
--- a/csit/metrics/dashboards/sla-metrics.json
+++ b/csit/metrics/dashboards/sla-metrics.json
@@ -23,18 +23,114 @@
   "fiscalYearStartMonth": 0,
   "gnetId": 14430,
   "graphTooltip": 0,
-  "id": 3,
-  "iteration": 1671098389991,
+  "id": 5,
+  "iteration": 1673087127172,
   "links": [],
   "liveNow": false,
   "panels": [
     {
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "line",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          }
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 0
+      },
+      "id": 135,
+      "options": {
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "single"
+        }
+      },
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "histogram_quantile(0.95, sum(rate(http_request_duration_seconds_bucket[5m])) by (le))",
+          "interval": "",
+          "legendFormat": "",
+          "refId": "A"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "",
+          "hide": false,
+          "interval": "",
+          "legendFormat": "",
+          "refId": "B"
+        }
+      ],
+      "title": "Panel Title",
+      "type": "timeseries"
+    },
+    {
       "collapsed": false,
       "gridPos": {
         "h": 1,
         "w": 24,
         "x": 0,
-        "y": 0
+        "y": 8
       },
       "id": 54,
       "panels": [],
@@ -80,7 +176,7 @@
         "h": 6,
         "w": 6,
         "x": 0,
-        "y": 1
+        "y": 9
       },
       "id": 52,
       "links": [],
@@ -158,7 +254,7 @@
         "h": 6,
         "w": 6,
         "x": 6,
-        "y": 1
+        "y": 9
       },
       "id": 56,
       "links": [],
@@ -185,7 +281,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "process_start_time_seconds{job=~\"api-metrics|acm-metrics|pap-metrics\"}*1000",
+          "expr": "process_start_time_seconds{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\"}*1000",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 2,
@@ -243,7 +339,7 @@
         "h": 6,
         "w": 6,
         "x": 0,
-        "y": 7
+        "y": 15
       },
       "id": 58,
       "links": [],
@@ -268,7 +364,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "sum(jvm_memory_used_bytes{job=~\"api-metrics|acm-metrics|pap-metrics\", area=\"heap\"})*100/sum(jvm_memory_max_bytes{job=~\"api-metrics|acm-metrics|pap-metrics\", area=\"heap\"})",
+          "expr": "sum(jvm_memory_used_bytes{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\", area=\"heap\"})*100/sum(jvm_memory_max_bytes{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\", area=\"heap\"})",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -335,7 +431,7 @@
         "h": 6,
         "w": 6,
         "x": 6,
-        "y": 7
+        "y": 15
       },
       "id": 60,
       "links": [],
@@ -360,7 +456,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "sum(jvm_memory_used_bytes{job=~\"api-metrics|acm-metrics|pap-metrics\", area=\"nonheap\"})*100/sum(jvm_memory_max_bytes{job=~\"api-metrics|acm-metrics|pap-metrics\", area=\"nonheap\"})",
+          "expr": "sum(jvm_memory_used_bytes{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\", area=\"nonheap\"})*100/sum(jvm_memory_max_bytes{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\", area=\"nonheap\"})",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 2,
@@ -429,7 +525,7 @@
         "h": 8,
         "w": 12,
         "x": 0,
-        "y": 13
+        "y": 21
       },
       "id": 96,
       "links": [],
@@ -456,7 +552,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "avg(system_load_average_1m{job=~\"api-metrics|acm-metrics|pap-metrics\"})",
+          "expr": "avg(system_load_average_1m{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\"})",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -469,7 +565,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "avg(system_cpu_count{job=~\"api-metrics|acm-metrics|pap-metrics\"})",
+          "expr": "avg(system_cpu_count{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\"})",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -537,7 +633,7 @@
         "h": 7,
         "w": 12,
         "x": 0,
-        "y": 21
+        "y": 29
       },
       "id": 95,
       "links": [],
@@ -564,7 +660,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "avg(system_cpu_usage{job=~\"api-metrics|acm-metrics|pap-metrics\"})",
+          "expr": "avg(system_cpu_usage{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\"})",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -577,7 +673,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "avg(process_cpu_usage{job=~\"api-metrics|acm-metrics|pap-metrics\"})",
+          "expr": "avg(process_cpu_usage{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\"})",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -644,7 +740,7 @@
         "h": 8,
         "w": 12,
         "x": 0,
-        "y": 28
+        "y": 36
       },
       "id": 113,
       "options": {
@@ -668,7 +764,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "avg by (id) (jvm_memory_used_bytes{job=~\"api-metrics|acm-metrics|pap-metrics\"})",
+          "expr": "avg by (id) (jvm_memory_used_bytes{job=~\"api-metrics|acm-metrics|pap-metrics|apex-pdp-metrics\"})",
           "interval": "",
           "legendFormat": "{{id}}",
           "refId": "A"
@@ -683,7 +779,7 @@
         "h": 1,
         "w": 24,
         "x": 0,
-        "y": 36
+        "y": 44
       },
       "id": 18,
       "panels": [],
@@ -747,7 +843,7 @@
         "h": 10,
         "w": 18,
         "x": 0,
-        "y": 37
+        "y": 45
       },
       "id": 4,
       "links": [],
@@ -772,7 +868,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "rate(http_server_requests_seconds_count[5m])",
+          "expr": "rate(http_server_requests_seconds_count{uri!~\".*metrics.*|.*prometheus.*\"}[5m])",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -840,7 +936,7 @@
         "h": 10,
         "w": 18,
         "x": 0,
-        "y": 47
+        "y": 55
       },
       "id": 2,
       "links": [],
@@ -866,7 +962,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "irate(http_server_requests_seconds_sum{job=~\"api-metrics|acm-metrics|pap-metrics\", exception=\"None\", uri!~\".*metrics.*\"}[5m]) / irate(http_server_requests_seconds_count{job=~\"api-metrics|acm-metrics|pap-metrics\", exception=\"None\", uri!~\".*metrics.*\"}[5m])",
+          "expr": "irate(http_server_requests_seconds_sum{job=~\"api-metrics|acm-metrics|pap-metrics\", exception=\"None\", uri!~\".*metrics.*|.*prometheus.*\"}[5m]) / irate(http_server_requests_seconds_count{job=~\"api-metrics|acm-metrics|pap-metrics\", exception=\"None\", uri!~\".*metrics.*|.*prometheus.*\"}[5m])",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -934,7 +1030,7 @@
         "h": 8,
         "w": 13,
         "x": 0,
-        "y": 57
+        "y": 65
       },
       "id": 111,
       "links": [],
@@ -959,7 +1055,7 @@
             "uid": "dkSf71fnz"
           },
           "exemplar": true,
-          "expr": "http_server_requests_seconds_max{job=~\"api-metrics|acm-metrics|pap-metrics\", uri!~\".*metrics.*\"}",
+          "expr": "http_server_requests_seconds_max{job=~\"api-metrics|acm-metrics|pap-metrics\", uri!~\".*metrics.*|.*prometheus.*\"}",
           "format": "time_series",
           "interval": "",
           "intervalFactor": 1,
@@ -976,7 +1072,7 @@
         "h": 1,
         "w": 24,
         "x": 0,
-        "y": 65
+        "y": 73
       },
       "id": 115,
       "panels": [],
@@ -1008,7 +1104,7 @@
         "h": 6,
         "w": 3,
         "x": 0,
-        "y": 66
+        "y": 74
       },
       "id": 117,
       "options": {
@@ -1067,7 +1163,7 @@
         "h": 6,
         "w": 3,
         "x": 3,
-        "y": 66
+        "y": 74
       },
       "id": 118,
       "options": {
@@ -1126,7 +1222,7 @@
         "h": 6,
         "w": 3,
         "x": 6,
-        "y": 66
+        "y": 74
       },
       "id": 119,
       "options": {
@@ -1185,7 +1281,7 @@
         "h": 6,
         "w": 3,
         "x": 9,
-        "y": 66
+        "y": 74
       },
       "id": 120,
       "options": {
@@ -1220,6 +1316,90 @@
       "type": "stat"
     },
     {
+      "description": "Policy Executions Total for APEX-PDP",
+      "fieldConfig": {
+        "defaults": {
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 80
+      },
+      "id": 128,
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "auto",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "textMode": "auto"
+      },
+      "pluginVersion": "8.3.4",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "pdpa_policy_executions_total{job=\"apex-pdp-metrics\", status=\"SUCCESS\"}",
+          "interval": "",
+          "legendFormat": "Successful",
+          "refId": "A"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "pdpa_policy_executions_total{job=\"apex-pdp-metrics\", status=\"FAIL\"}",
+          "hide": false,
+          "interval": "",
+          "legendFormat": "Failed",
+          "refId": "B"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "pdpa_policy_executions_total{job=\"apex-pdp-metrics\", status=\"TOTAL\"}",
+          "hide": false,
+          "interval": "",
+          "legendFormat": "Total",
+          "refId": "C"
+        }
+      ],
+      "title": "Policy Executions",
+      "type": "stat"
+    },
+    {
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -1275,7 +1455,7 @@
         "h": 8,
         "w": 12,
         "x": 0,
-        "y": 72
+        "y": 88
       },
       "id": 122,
       "options": {
@@ -1307,6 +1487,347 @@
       ],
       "title": "Deployment timing",
       "type": "timeseries"
+    },
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 96
+      },
+      "id": 124,
+      "panels": [],
+      "title": "Apex-PDP",
+      "type": "row"
+    },
+    {
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "fixed"
+          },
+          "mappings": [
+            {
+              "options": {
+                "0": {
+                  "color": "red",
+                  "index": 0,
+                  "text": "Undefined"
+                },
+                "1": {
+                  "color": "#ffffff",
+                  "index": 1,
+                  "text": "Stopped"
+                },
+                "2": {
+                  "color": "blue",
+                  "index": 2,
+                  "text": "Ready"
+                },
+                "3": {
+                  "color": "green",
+                  "index": 3,
+                  "text": "Running"
+                },
+                "4": {
+                  "color": "yellow",
+                  "index": 4,
+                  "text": "Stopping"
+                }
+              },
+              "type": "value"
+            }
+          ],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 5,
+        "x": 0,
+        "y": 97
+      },
+      "id": 126,
+      "options": {
+        "colorMode": "value",
+        "graphMode": "area",
+        "justifyMode": "auto",
+        "orientation": "horizontal",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "text": {
+          "titleSize": 20,
+          "valueSize": 20
+        },
+        "textMode": "auto"
+      },
+      "pluginVersion": "8.3.4",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "pdpa_engine_state{job=\"apex-pdp-metrics\"}",
+          "interval": "",
+          "legendFormat": "{{ engine_instance_id }}",
+          "refId": "A"
+        }
+      ],
+      "title": "Engines Stats",
+      "type": "stat"
+    },
+    {
+      "description": "Uptime per engine for Policy APEX-PDP application calculated in days",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "decimals": 0,
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unit": "none"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 5,
+        "x": 5,
+        "y": 97
+      },
+      "id": 130,
+      "options": {
+        "colorMode": "none",
+        "graphMode": "none",
+        "justifyMode": "auto",
+        "orientation": "horizontal",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "text": {
+          "titleSize": 25,
+          "valueSize": 25
+        },
+        "textMode": "auto"
+      },
+      "pluginVersion": "8.3.4",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "avg by(engine_instance_id)((time() - pdpa_engine_last_start_timestamp_epoch{job=\"apex-pdp-metrics\"}/1000)/86400)",
+          "format": "time_series",
+          "instant": false,
+          "interval": "",
+          "legendFormat": "{{ engine_instance_id }}",
+          "refId": "A"
+        }
+      ],
+      "title": "Engine Uptime (days)",
+      "transformations": [],
+      "type": "stat"
+    },
+    {
+      "description": "Number of APEX event execution counter per engine thread",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "line",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unit": "ops"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 105
+      },
+      "id": 132,
+      "options": {
+        "legend": {
+          "calcs": [
+            "lastNotNull",
+            "min",
+            "max"
+          ],
+          "displayMode": "table",
+          "placement": "right"
+        },
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
+      },
+      "pluginVersion": "8.4.5",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "rate(pdpa_engine_average_execution_time_seconds{job=\"apex-pdp-metrics\"}[5m])",
+          "format": "time_series",
+          "interval": "",
+          "legendFormat": "{{ engine_instance_id }}",
+          "refId": "A"
+        }
+      ],
+      "title": "Avg Engine Execution Time",
+      "type": "timeseries"
+    },
+    {
+      "description": "Number of APEX event execution counter per engine thread",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            }
+          },
+          "mappings": [],
+          "unit": "short"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 113
+      },
+      "id": 133,
+      "options": {
+        "displayLabels": [
+          "percent"
+        ],
+        "legend": {
+          "displayMode": "table",
+          "placement": "right",
+          "values": [
+            "value"
+          ]
+        },
+        "pieType": "donut",
+        "reduceOptions": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "fields": "",
+          "values": false
+        },
+        "tooltip": {
+          "mode": "single"
+        }
+      },
+      "pluginVersion": "8.4.5",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "dkSf71fnz"
+          },
+          "exemplar": true,
+          "expr": "pdpa_engine_event_executions{job=\"apex-pdp-metrics\"}",
+          "format": "time_series",
+          "interval": "",
+          "legendFormat": "{{ engine_instance_id }}",
+          "refId": "A"
+        }
+      ],
+      "title": "Engine Executions Count",
+      "type": "piechart"
     }
   ],
   "refresh": "10s",
@@ -1378,6 +1899,6 @@
   "timezone": "",
   "title": "SLA's Metrics",
   "uid": "ySoel0jnk",
-  "version": 3,
+  "version": 4,
   "weekStart": ""
 }
\ No newline at end of file