fix(logging): suppress successful health-probe access logs across all services

dashboard/api/Dockerfile

@@ -36,4 +36,4 @@ EXPOSE 5000
 HEALTHCHECK --interval=30s --timeout=3s --start-period=10s --retries=3 \
     CMD curl --fail http://localhost:5000/health || exit 1
 
-CMD ["gunicorn", "--bind", "0.0.0.0:5000", "--workers", "1", "--threads", "2", "--access-logfile", "-", "--error-logfile", "-", "app:create_app()"]
+CMD ["gunicorn", "--bind", "0.0.0.0:5000", "--workers", "1", "--threads", "2", "--access-logfile", "-", "--error-logfile", "-", "--logger-class", "gunicorn_logging.HealthCheckFilter", "app:create_app()"]

dashboard/api/gunicorn_logging.py

@@ -0,0 +1,17 @@
+"""Gunicorn logging helpers."""
+
+from datetime import timedelta
+
+from gunicorn.glogging import Logger
+
+
+class HealthCheckFilter(Logger):
+    """Suppress successful health-check access logs."""
+
+    def access(self, resp: object, req: object, environ: dict[str, str], request_time: timedelta) -> None:
+        """Log non-health requests and failed health probes."""
+        path = environ.get("PATH_INFO", "").rstrip("/")
+        status = str(getattr(resp, "status", "")).partition(" ")[0]
+        if path.endswith("/health") and status == "200":
+            return
+        super().access(resp, req, environ, request_time)

dashboard/api/tests/unit/test_container_config.py

@@ -0,0 +1,12 @@
+"""Tests for dashboard API container runtime configuration."""
+
+from pathlib import Path
+
+
+def test_gunicorn_suppresses_successful_health_access_logs() -> None:
+    """Routine successful /health probes should not flood stdout logs."""
+    dockerfile = Path(__file__).parents[2] / "Dockerfile"
+    content = dockerfile.read_text(encoding="utf-8")
+
+    assert "--logger-class" in content
+    assert "gunicorn_logging.HealthCheckFilter" in content

dashboard/auth/Dockerfile

@@ -27,4 +27,4 @@ EXPOSE 5001
 HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
     CMD wget --spider --quiet http://localhost:5001/health || exit 1
 
-CMD ["gunicorn", "--bind", "0.0.0.0:5001", "--workers", "1", "--access-logfile", "-", "--error-logfile", "-", "auth:app"]
+CMD ["gunicorn", "--bind", "0.0.0.0:5001", "--workers", "1", "--access-logfile", "-", "--error-logfile", "-", "--logger-class", "gunicorn_logging.HealthCheckFilter", "auth:app"]

dashboard/auth/auth.py

@@ -16,7 +16,6 @@
 register_error_handlers(app)
 
 logger = logging.getLogger(__name__)
-_first_health_logged = False
 
 
 # Environment/config
@@ -145,10 +144,6 @@ def health() -> tuple[str, int]:
     Returns:
         tuple[str, int]: A plain-text OK body and a 200 status code.
     """
-    global _first_health_logged  # ruff:ignore[global-statement]
-    if not _first_health_logged:
-        logger.info("auth first health response served")
-        _first_health_logged = True
     return "OK", HTTPStatus.OK
 
 

dashboard/auth/gunicorn_logging.py

@@ -0,0 +1,17 @@
+"""Gunicorn logging helpers."""
+
+from datetime import timedelta
+
+from gunicorn.glogging import Logger
+
+
+class HealthCheckFilter(Logger):
+    """Suppress successful health-check access logs."""
+
+    def access(self, resp: object, req: object, environ: dict[str, str], request_time: timedelta) -> None:
+        """Log non-health requests and failed health probes."""
+        path = environ.get("PATH_INFO", "").rstrip("/")
+        status = str(getattr(resp, "status", "")).partition(" ")[0]
+        if path.endswith("/health") and status == "200":
+            return
+        super().access(resp, req, environ, request_time)

dashboard/auth/tests/test_auth.py

@@ -288,17 +288,3 @@ def test_login_token_requires_token_query_parameter(self, client: FlaskClient) -
         response = client.get("/login-token")
 
         assert response.status_code == HTTPStatus.BAD_REQUEST
-
-
-def test_health_logs_first_health_once(client: FlaskClient, caplog) -> None:
-    """The first auth health response should be visible in startup diagnostics once."""
-    import auth
-
-    auth._first_health_logged = False
-    caplog.set_level("INFO", logger="auth")
-
-    client.get("/health")
-    client.get("/health")
-
-    messages = [record.getMessage() for record in caplog.records]
-    assert messages.count("auth first health response served") == 1

dashboard/auth/tests/test_container_config.py

@@ -0,0 +1,12 @@
+"""Tests for auth container runtime configuration."""
+
+from pathlib import Path
+
+
+def test_gunicorn_suppresses_successful_health_access_logs() -> None:
+    """Routine successful /health probes should not flood stdout logs."""
+    dockerfile = Path(__file__).parents[1] / "Dockerfile"
+    content = dockerfile.read_text(encoding="utf-8")
+
+    assert "--logger-class" in content
+    assert "gunicorn_logging.HealthCheckFilter" in content

mdrun-api/gunicorn_logging.py

@@ -1,37 +1,17 @@
 """Gunicorn logging helpers."""
 
 from datetime import timedelta
-from typing import Any
 
 from gunicorn.glogging import Logger
 
-HTTP_SUCCESS_MIN = 200
-HTTP_REDIRECT_MIN = 300
-
 
 class HealthCheckFilter(Logger):
     """Suppress successful health-check access logs."""
 
-    def access(self, resp: Any, req: Any, environ: dict[str, Any], request_time: timedelta) -> None:  # ruff:ignore[any-type]
+    def access(self, resp: object, req: object, environ: dict[str, str], request_time: timedelta) -> None:
         """Log non-health requests and failed health probes."""
-        status_code = _get_status_code(resp)
-        path = environ.get("PATH_INFO", "")
-        if path == "/api/health" and status_code is not None and HTTP_SUCCESS_MIN <= status_code < HTTP_REDIRECT_MIN:
+        path = environ.get("PATH_INFO", "").rstrip("/")
+        status = str(getattr(resp, "status", "")).partition(" ")[0]
+        if path.endswith("/health") and status == "200":
             return
         super().access(resp, req, environ, request_time)
-
-
-def _get_status_code(resp: Any) -> int | None:  # ruff:ignore[any-type]
-    status_int = getattr(resp, "status_int", None)
-    if isinstance(status_int, int):
-        return status_int
-
-    status = getattr(resp, "status", None)
-    if isinstance(status, int):
-        return status
-    if isinstance(status, str):
-        code = status.split(maxsplit=1)[0]
-        if code.isdigit():
-            return int(code)
-
-    return None

tuner/api/access_logging.py

@@ -0,0 +1,25 @@
+"""
+Uvicorn access log filtering.
+
+The uvicorn access logger emits one record per request as
+``logger.info(fmt, (client_addr, method, full_path, http_version, status_code))``;
+``AccessFormatter.formatMessage`` unpacks that 5-tuple from ``record.args``.
+"""
+
+import logging
+
+
+class HealthCheckFilter(logging.Filter):
+    """Suppress successful health-check access logs."""
+
+    def filter(self, record: logging.LogRecord) -> bool:
+        """Return False to drop successful health probes; keep all else."""
+        args = record.args
+        if not isinstance(args, tuple):
+            return True
+        try:
+            path = str(args[2]).rstrip("/")
+            status = str(args[4])
+        except IndexError:
+            return True
+        return not (path.endswith("/health") and status == "200")

tuner/api/start.py

@@ -5,6 +5,7 @@
 import uvicorn
 from alembic import command
 from alembic.config import Config
+from uvicorn.config import LOGGING_CONFIG
 
 logging.basicConfig(
     level=logging.INFO,
@@ -13,6 +14,10 @@
 )
 logger = logging.getLogger(__name__)
 
+# Suppress successful health-probe access logs for parity with the gunicorn services.
+LOGGING_CONFIG.setdefault("filters", {})["healthcheck"] = {"()": "api.access_logging.HealthCheckFilter"}
+LOGGING_CONFIG["loggers"]["uvicorn.access"].setdefault("filters", []).append("healthcheck")
+
 _REQUIRED_ENV = ("TUNER_USER", "TUNER_PASSWORD", "COST_CPU_CORE_HOUR", "COST_GPU_HOUR", "COST_GB_RAM_HOUR")
 
 
@@ -35,7 +40,7 @@ def main() -> None:
     from api.main import app  # ruff: ignore[import-outside-top-level] — intentional: import after migrations complete
 
     try:
-        uvicorn.run(app, host="0.0.0.0", port=8000)
+        uvicorn.run(app, host="0.0.0.0", port=8000, log_config=LOGGING_CONFIG)
     except Exception:
         logger.exception("Unexpected error while starting.")
         sys.exit(1)

tuner/tests/test_container_config.py

@@ -7,3 +7,11 @@ def test_ray_runtime_workdir_excludes_api_virtual_environment() -> None:
     assert "COPY --chown=app:app tuner/api/ tuner-runtime/api/" in dockerfile
     assert "RUNTIME_WORKDIR=/app/tuner-runtime" in dockerfile
     assert 'PYTHONPATH="/app/tuner-runtime"' in dockerfile
+
+
+def test_uvicorn_suppresses_successful_health_access_logs() -> None:
+    """Routine successful /api/health probes should not flood stdout logs."""
+    start_py = (Path(__file__).parents[1] / "api" / "start.py").read_text(encoding="utf-8")
+
+    assert "api.access_logging.HealthCheckFilter" in start_py
+    assert "log_config=LOGGING_CONFIG" in start_py