refactor: logging

Signed-off-by: thxCode <thxcode0824@gmail.com>

Author: thxCode

gpustack_runtime/deployer/docker.py

@@ -273,7 +273,7 @@ def is_supported() -> bool:
                 supported = client.ping()
             except docker.errors.APIError:
                 if logger.isEnabledFor(logging.DEBUG):
-                    logger.exception("Docker ping failed")
+                    logger.exception("Failed to connect to Docker API server")
 
         return supported
 

gpustack_runtime/deployer/kuberentes.py

@@ -280,10 +280,10 @@ def is_supported() -> bool:
                 version_api = kubernetes.client.VersionApi(client)
                 version_info = version_api.get_code()
                 supported = version_info is not None
-            except kubernetes.client.exceptions.ApiException:
-                if logger.isEnabledFor(logging.DEBUG):
-                    logger.exception("Failed to get Kubernetes version")
-            except urllib3.exceptions.MaxRetryError:
+            except (
+                urllib3.exceptions.MaxRetryError,
+                kubernetes.client.exceptions.ApiException,
+            ):
                 if logger.isEnabledFor(logging.DEBUG):
                     logger.exception("Failed to connect to Kubernetes API server")
 

gpustack_runtime/detector/ascend.py

@@ -10,6 +10,7 @@
 from .__utils__ import PCIDevice, get_brief_version, get_pci_devices, get_utilization
 
 logger = logging.getLogger(__name__)
+slogger = logger.getChild("internal")
 
 
 class AscendDetector(Detector):
@@ -261,8 +262,8 @@ def _get_device_roce_network_info(
             pydcmi.DCMI_PORT_TYPE_ROCE_PORT,
         )
     except pydcmi.DCMIError:
-        if logger.isEnabledFor(logging.DEBUG):
-            logger.exception("Failed to get device roce network info")
+        if slogger.isEnabledFor(logging.DEBUG):
+            slogger.exception("Failed to get device roce network info")
 
     return ip, mask, gateway
 
@@ -288,8 +289,8 @@ def _get_device_virtual_info(
             c_vdev_query_stru,
         )
     except pydcmi.DCMIError:
-        if logger.isEnabledFor(logging.DEBUG):
-            logger.exception("Failed to get device virtual info")
+        if slogger.isEnabledFor(logging.DEBUG):
+            slogger.exception("Failed to get device virtual info")
     else:
         return c_vdev_query_stru
 

gpustack_runtime/logging.py

@@ -27,11 +27,12 @@ def _parse_module_levels(level_str: str) -> dict[str, int]:
     Parse the GPUSTACK_RUNTIME_LOG_LEVEL environment variable to determine module-specific log levels.
 
     Examples:
-        - "DEBUG"                                         # All modules at DEBUG
-        - "runtime.module_a:DEBUG"                        # Only module_a module at DEBUG, other modules at INFO
-        - "module_a:DEBUG"                                # Same as above
-        - "runtime.module_a:DEBUG;runtime.module_b:INFO"  # Multiple modules
-        - "ERROR;runtime.module_a:DEBUG"                  # All modules at ERROR, only module_a module at DEBUG
+        - "DEBUG"                                                           # All modules at DEBUG
+        - "gpustack_runtime.module_a:DEBUG"                                 # Only module_a module at DEBUG, other modules at INFO
+        - "module_a:DEBUG"                                                  # Same as above
+        - "module_a=DEBUG"                                                  # Using '=' instead of ':'
+        - "gpustack_runtime.module_a:DEBUG;gpustack_runtime.module_b:INFO"  # Multiple modules
+        - "ERROR;runtime.module_a:DEBUG"                                    # All modules at ERROR, only module_a module at DEBUG
 
     """
     module_levels: dict[str, int] = {}  # {"module_name": log_level}
@@ -48,14 +49,16 @@ def _parse_module_levels(level_str: str) -> dict[str, int]:
         module = ""
         level = ""
         if ":" in p:
-            module, level = p.split(":")
+            module, level = p.split(":", 1)
+        elif "=" in p:
+            module, level = p.split("=", 1)
 
         level = level.upper()
         if level not in levels:
             continue
 
         module = module.strip()
-        module = module.replace("runtime.", "")
+        module = module.replace(f"{__package__}.", "")
         module = module.replace("/", ".").strip(".")
 
         module_levels[module] = getattr(logging, level)