ING-1419: Generate graphs from the spectroperf metrics

Author: Westwooo

.github/workflows/load_test.yml

@@ -1,5 +1,6 @@
 name: Run Load Tests
 permissions:
+  pull-requests: write
   contents: read
   packages: read
 
@@ -99,6 +100,41 @@ jobs:
         run: |
           cat ./gateway/metrics.json
 
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11' # Specify your version
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install numpy matplotlib
+
+      - name: Generate graphs from metrics
+        run: |
+          mv ./direct/metrics.json ./direct.json
+          mv ./gateway/metrics.json ./gateway.json
+          python plot_metrics.py 50
+          python plot_metrics.py 95
+
+      - name: Upload Graphs as Artifacts
+        id: upload-graphs
+        uses: actions/upload-artifact@v4
+        with:
+          name: performance-graphs-${{ matrix.server }}
+          path: |
+            metrics_p50.png
+            metrics_p95.png
+
+      - name: Post Graphs to PR
+        uses: thollander/actions-comment-pull-request@v2
+        if: github.event_name == 'pull_request'
+        with:
+          comment_tag: perf-report-${{ matrix.server }}
+          message: |
+            ### 📊 Performance Metrics Report ${{ matrix.server }}
+            [View Performance Graphs](${{ steps.upload-graphs.outputs.artifact-url }})
+
       - name: Collect couchbase logs
         timeout-minutes: 10
         if: failure()

plot_metrics.py

@@ -0,0 +1,77 @@
+import sys
+import json
+import numpy as np
+import matplotlib
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+
+PERCENTILE_MAP = {
+    50: "fiftieth",
+    95: "ninetyFifth",
+    98: "ninetyEighth",
+    99: "ninetyNinth",
+}
+
+if len(sys.argv) != 2 or not sys.argv[1].isdigit() or int(sys.argv[1]) not in PERCENTILE_MAP:
+    print(f"Usage: python plot_metrics.py <percentile>")
+    print(f"Valid percentiles: {sorted(PERCENTILE_MAP)}")
+    sys.exit(1)
+
+percentile = int(sys.argv[1])
+percentile_key = PERCENTILE_MAP[percentile]
+
+with open("direct.json") as f:
+    direct = json.load(f)
+
+with open("gateway.json") as f:
+    gateway = json.load(f)
+
+def extract_data(data, key):
+    ops = {}
+    for entry in data:
+        num_users = entry["numUsers"]
+        for op in entry["operations"]:
+            name = op["name"]
+            if name not in ops:
+                ops[name] = {"numUsers": [], "latency": []}
+            ops[name]["numUsers"].append(num_users)
+            ops[name]["latency"].append(op["latencyPercentiles"][key])
+    return ops
+
+def smooth(x, y):
+    x = np.array(x)
+    y = np.array(y)
+    coeffs = np.polyfit(x, y, deg=min(2, len(x) - 1))
+    x_smooth = np.linspace(x.min(), x.max(), 300)
+    y_smooth = np.polyval(coeffs, x_smooth)
+    return x_smooth, y_smooth
+
+direct_ops = extract_data(direct, percentile_key)
+gateway_ops = extract_data(gateway, percentile_key)
+
+operation_names = list(direct_ops.keys())
+
+fig, axes = plt.subplots(1, len(operation_names), figsize=(5 * len(operation_names), 5))
+fig.suptitle(f"{percentile}th Percentile Latency by Operation", fontsize=14, fontweight="bold", y=1.02)
+
+for ax, name in zip(axes, operation_names):
+    for ops, label, marker, color in [
+        (direct_ops, "Direct", "o", "tab:blue"),
+        (gateway_ops, "Gateway", "s", "tab:orange"),
+    ]:
+        x = ops[name]["numUsers"]
+        y = ops[name]["latency"]
+        x_smooth, y_smooth = smooth(x, y)
+        ax.plot(x_smooth, y_smooth, color=color, linewidth=1.5)
+        ax.plot(x, y, marker=marker, linestyle="None", color=color, label=label)
+
+    ax.set_title(name)
+    ax.set_xlabel("Number of Users")
+    ax.set_ylabel(f"Latency p{percentile} (ms)")
+    ax.legend()
+    ax.grid(True)
+
+plt.tight_layout()
+output_file = f"metrics_p{percentile}.png"
+plt.savefig(output_file, dpi=150, bbox_inches="tight")
+print(f"Saved {output_file}")