20260623 增加RP差异趋势面板

monitor/grafana/dashboards/ours-rp-inter-rp.json

@@ -452,6 +452,114 @@
         }
       ]
     },
+    {
+      "id": 15,
+      "title": "VRP Diff Trend by Class",
+      "type": "timeseries",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "Prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 0,
+        "y": 28
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "none",
+          "min": 0,
+          "decimals": 0
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "none"
+        }
+      },
+      "targets": [
+        {
+          "expr": "inter_rp_vrps_diff_by_class{exported_instance=~\".*inter-rp\",left=\"ours-rp\",right=\"routinator\",class=\"total\"}",
+          "legendFormat": "total diff",
+          "refId": "A"
+        },
+        {
+          "expr": "inter_rp_vrps_diff_by_class{exported_instance=~\".*inter-rp\",left=\"ours-rp\",right=\"routinator\",class=\"only_ours\"}",
+          "legendFormat": "only ours",
+          "refId": "B"
+        },
+        {
+          "expr": "inter_rp_vrps_diff_by_class{exported_instance=~\".*inter-rp\",left=\"ours-rp\",right=\"routinator\",class=\"only_routinator\"}",
+          "legendFormat": "only routinator",
+          "refId": "C"
+        }
+      ]
+    },
+    {
+      "id": 16,
+      "title": "VAP Diff Trend by Class",
+      "type": "timeseries",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "Prometheus"
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 28
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "none",
+          "min": 0,
+          "decimals": 0
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "multi",
+          "sort": "none"
+        }
+      },
+      "targets": [
+        {
+          "expr": "inter_rp_vaps_diff_by_class{exported_instance=~\".*inter-rp\",left=\"ours-rp\",right=\"routinator\",class=\"total\"}",
+          "legendFormat": "total diff",
+          "refId": "A"
+        },
+        {
+          "expr": "inter_rp_vaps_diff_by_class{exported_instance=~\".*inter-rp\",left=\"ours-rp\",right=\"routinator\",class=\"only_ours\"}",
+          "legendFormat": "only ours",
+          "refId": "B"
+        },
+        {
+          "expr": "inter_rp_vaps_diff_by_class{exported_instance=~\".*inter-rp\",left=\"ours-rp\",right=\"routinator\",class=\"only_routinator\"}",
+          "legendFormat": "only routinator",
+          "refId": "C"
+        }
+      ]
+    },
     {
       "id": 11,
       "title": "Artifact Age by RP",
@@ -464,7 +572,7 @@
         "h": 8,
         "w": 24,
         "x": 0,
-        "y": 28
+        "y": 36
       },
       "fieldConfig": {
         "defaults": {
@@ -507,7 +615,7 @@
         "h": 8,
         "w": 12,
         "x": 0,
-        "y": 36
+        "y": 44
       },
       "fieldConfig": {
         "defaults": {
@@ -551,7 +659,7 @@
         "h": 8,
         "w": 12,
         "x": 12,
-        "y": 36
+        "y": 44
       },
       "fieldConfig": {
         "defaults": {
@@ -595,7 +703,7 @@
         "h": 9,
         "w": 24,
         "x": 0,
-        "y": 44
+        "y": 52
       },
       "fieldConfig": {
         "defaults": {

scripts/inter_rp/inter_rp_ours_routinator_exporter.py

@@ -20,6 +20,7 @@ SCAN_TTL = float(os.environ.get("INTER_RP_SCAN_TTL_SECONDS", "10"))
 _cache_lock = threading.Lock()
 _cache = {"deadline": 0.0, "metrics": "", "status": {}}
 _count_cache = {}
+_set_cache = {}
 
 def unix_now():
     return time.time()
@@ -42,13 +43,13 @@ def latest_ours_run():
     candidates = sorted(p for p in runs.glob("run_*") if (p / "run-summary.json").exists())
     return candidates[-1] if candidates else None
 
-def count_unique_csv(path, cols):
+def unique_csv_set(path, cols):
     if not path.exists():
         return None
     stat = path.stat()
     key = (str(path), stat.st_mtime_ns, stat.st_size, cols)
-    if key in _count_cache:
+    if key in _set_cache:
-        return _count_cache[key]
+        return _set_cache[key]
     seen = set()
     with open(path, "r", encoding="utf-8", newline="") as handle:
         reader = csv.reader(handle)
@@ -62,10 +63,15 @@ def count_unique_csv(path, cols):
             if len(row) < cols:
                 continue
             seen.add(tuple(cell.strip() for cell in row[:cols]))
-    value = len(seen)
+    _set_cache.clear()
-    _count_cache.clear()
+    _set_cache[key] = seen
-    _count_cache[key] = value
+    return seen
-    return value
+
+def count_unique_csv(path, cols):
+    values = unique_csv_set(path, cols)
+    if values is None:
+        return None
+    return len(values)
 
 def metric_line(name, labels, value):
     label_text = ",".join(f'{k}="{str(v).replace(chr(92), chr(92)+chr(92)).replace(chr(34), chr(92)+chr(34))}"' for k, v in labels.items())
@@ -102,8 +108,10 @@ def sample_ours(now, errors):
     if rss is not None:
         sample["max"]["parent"] = int(rss) * 1024
         sample["max"]["aggregate_peak"] = int(rss) * 1024
-    sample["vrps"] = count_unique_csv(run_dir / "vrps.csv", 3)
+    sample["vrp_set"] = unique_csv_set(run_dir / "vrps.csv", 3)
-    sample["vaps"] = count_unique_csv(run_dir / "vaps.csv", 2)
+    sample["vap_set"] = unique_csv_set(run_dir / "vaps.csv", 2)
+    sample["vrps"] = len(sample["vrp_set"]) if sample["vrp_set"] is not None else None
+    sample["vaps"] = len(sample["vap_set"]) if sample["vap_set"] is not None else None
     return sample
 
 def sample_routinator(now, errors):
@@ -135,8 +143,10 @@ def sample_routinator(now, errors):
     for label, key in [("parent", "parent"), ("child_max", "childMax"), ("aggregate_peak", "aggregatePeak")]:
         if max_rss.get(key) is not None:
             sample["max"][label] = int(max_rss[key]) * 1024
-    sample["vrps"] = count_unique_csv(latest / "vrps.csv", 3)
+    sample["vrp_set"] = unique_csv_set(latest / "vrps.csv", 3)
-    sample["vaps"] = count_unique_csv(latest / "vaps.csv", 2)
+    sample["vap_set"] = unique_csv_set(latest / "vaps.csv", 2)
+    sample["vrps"] = len(sample["vrp_set"]) if sample["vrp_set"] is not None else None
+    sample["vaps"] = len(sample["vap_set"]) if sample["vap_set"] is not None else None
     return sample
 
 
@@ -384,10 +394,34 @@ def build_metrics():
             labels2["kind"] = kind
             out.append(metric_line("inter_rp_run_max_rss_bytes", labels2, value))
     ours, rout = by_rp.get("ours-rp", {}), by_rp.get("routinator", {})
-    if ours.get("vrps") is not None and rout.get("vrps") is not None:
+    if ours.get("vrp_set") is not None and rout.get("vrp_set") is not None:
-        out.append(metric_line("inter_rp_vrps_diff", {"instance": INSTANCE, "left": "ours-rp", "right": "routinator"}, abs(int(ours["vrps"]) - int(rout["vrps"]))))
+        only_ours = ours["vrp_set"] - rout["vrp_set"]
-    if ours.get("vaps") is not None and rout.get("vaps") is not None:
+        only_rout = rout["vrp_set"] - ours["vrp_set"]
-        out.append(metric_line("inter_rp_vaps_diff", {"instance": INSTANCE, "left": "ours-rp", "right": "routinator"}, abs(int(ours["vaps"]) - int(rout["vaps"]))))
+        total_diff = len(only_ours) + len(only_rout)
+        labels = {"instance": INSTANCE, "left": "ours-rp", "right": "routinator"}
+        out.append(metric_line("inter_rp_vrps_diff", labels, total_diff))
+        out.append(metric_line("inter_rp_vrps_diff_by_class", {**labels, "class": "total"}, total_diff))
+        out.append(metric_line("inter_rp_vrps_diff_by_class", {**labels, "class": "only_ours"}, len(only_ours)))
+        out.append(metric_line("inter_rp_vrps_diff_by_class", {**labels, "class": "only_routinator"}, len(only_rout)))
+    elif ours.get("vrps") is not None and rout.get("vrps") is not None:
+        labels = {"instance": INSTANCE, "left": "ours-rp", "right": "routinator"}
+        total_diff = abs(int(ours["vrps"]) - int(rout["vrps"]))
+        out.append(metric_line("inter_rp_vrps_diff", labels, total_diff))
+        out.append(metric_line("inter_rp_vrps_diff_by_class", {**labels, "class": "total"}, total_diff))
+    if ours.get("vap_set") is not None and rout.get("vap_set") is not None:
+        only_ours = ours["vap_set"] - rout["vap_set"]
+        only_rout = rout["vap_set"] - ours["vap_set"]
+        total_diff = len(only_ours) + len(only_rout)
+        labels = {"instance": INSTANCE, "left": "ours-rp", "right": "routinator"}
+        out.append(metric_line("inter_rp_vaps_diff", labels, total_diff))
+        out.append(metric_line("inter_rp_vaps_diff_by_class", {**labels, "class": "total"}, total_diff))
+        out.append(metric_line("inter_rp_vaps_diff_by_class", {**labels, "class": "only_ours"}, len(only_ours)))
+        out.append(metric_line("inter_rp_vaps_diff_by_class", {**labels, "class": "only_routinator"}, len(only_rout)))
+    elif ours.get("vaps") is not None and rout.get("vaps") is not None:
+        labels = {"instance": INSTANCE, "left": "ours-rp", "right": "routinator"}
+        total_diff = abs(int(ours["vaps"]) - int(rout["vaps"]))
+        out.append(metric_line("inter_rp_vaps_diff", labels, total_diff))
+        out.append(metric_line("inter_rp_vaps_diff_by_class", {**labels, "class": "total"}, total_diff))
     repo_diff = emit_repo_diff_metrics(out, errors)
     return "".join(out), {"errors": errors, "samples": samples, "sync": sync, "repoDiff": repo_diff}