Fix benchmark cache reset and log routines

Author: crimist

bench/.gitignore

@@ -1,3 +1,3 @@
 trakxbench
-
 __pycache__/
+results/

bench/README.md

@@ -35,7 +35,7 @@ The orchestrator will launch Trakx with:
 trakx --config bench/trakx.yaml start
 ```
 
-It also sets `TRAKX_CACHE` to a benchmark-only directory and deletes it between runs.
+It deletes the configured cache directory between runs to reset state.
 
 ## Orchestrated workflow (recommended)
 

bench/orchestrator_server.py

@@ -105,9 +105,43 @@ def wait_for_udp(host: str, port: int, timeout: float) -> bool:
     return False
 
 
-def clear_cache(cache_dir: str) -> None:
+def default_cache_dir() -> Optional[Path]:
+    xdg = os.environ.get("XDG_CACHE_HOME")
+    if xdg:
+        return Path(xdg) / "trakx"
+    home = Path.home()
+    if home:
+        return home / ".cache" / "trakx"
+    return None
+
+
+def read_cache_path(config_path: str, repo_root: Path) -> Optional[Path]:
+    try:
+        with open(config_path, "r", encoding="utf-8") as fh:
+            for line in fh:
+                stripped = line.strip()
+                if not stripped or stripped.startswith("#"):
+                    continue
+                if stripped.startswith("cache:"):
+                    value = stripped[len("cache:") :].strip()
+                    if not value:
+                        return default_cache_dir()
+                    value = value.strip("\"'")
+                    path = Path(value).expanduser()
+                    if not path.is_absolute():
+                        path = repo_root / path
+                    return path
+    except OSError:
+        return default_cache_dir()
+    return default_cache_dir()
+
+
+def clear_cache(cache_dir: Optional[Path]) -> None:
     if not cache_dir:
         return
+    cache_dir = cache_dir.resolve()
+    if cache_dir == Path("/"):
+        return
     shutil.rmtree(cache_dir, ignore_errors=True)
 
 
@@ -132,6 +166,8 @@ def main() -> int:
 
     listen_host, listen_port = parse_hostport(args.listen)
     env_base = os.environ.copy()
+    env_base.pop("TRAKX_CACHE", None)
+    cache_dir = read_cache_path(args.config, repo_root)
 
     server_sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
     server_sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
@@ -161,9 +197,7 @@ def main() -> int:
                 env["TRAKX_HTTP_ROUTINES"] = str(http_routines)
 
             _ = run_cmd([args.trakx_bin, "--config", args.config, "stop"], env=env)
-            cache_dir = env.get("TRAKX_CACHE")
-            if cache_dir:
-                clear_cache(cache_dir)
+            clear_cache(cache_dir)
 
             start = run_cmd([args.trakx_bin, "--config", args.config, "start"], env=env)
             if start.returncode != 0:

tracker/http/httptracker.go

@@ -55,7 +55,7 @@ func (tracker *Tracker) Serve(ip net.IP, port int, routines int) error {
 	if err != nil {
 		return errors.Wrap(err, "Failed to open TCP listen socket")
 	}
-	zap.L().Info("Serving HTTP tracker on", zap.String("address", listener.Addr().String()))
+	zap.L().Info("Serving HTTP tracker", zap.String("address", listener.Addr().String()), zap.Int("routines", routines))
 
 	// TODO: figure out what optimal number of goroutines is (benchmark)
 	// Going to need to write a tool that can simulate a large number of clients

tracker/udp/udptracker.go

@@ -66,7 +66,7 @@ func (tracker *Tracker) Serve(ip net.IP, port int, routines int) error {
 	if err != nil {
 		return errors.Wrap(err, "Failed to open UDP listen socket")
 	}
-	zap.L().Info("Serving UDP tracker on", zap.String("address", tracker.socket.LocalAddr().String()))
+	zap.L().Info("Serving UDP tracker", zap.String("address", tracker.socket.LocalAddr().String()), zap.Int("routines", routines))
 
 	// TODO: figure out what optimal number of goroutines is (benchmark)
 	// Going to need to write a tool that can simulate a large number of clients