Add task-specific log capturing to task diagnostics mechanism

TASK_DIAGNOSTICS and X-Task-Diagnostics can now capture all logs for
specific tasks, including DEBUG logs, regardless of whether DEBUG logs
are otherwise enabled for the system logger. Because the logs are
specific to the task, the logs are linear and not broken up by
unrelated logs from other services/workers/tasks.

Assisted By: Claude

closes #7214

Author: dralley

CHANGES/7214.feature

@@ -0,0 +1 @@
+Added a "logs" option to the TASK_DIAGNOSTICS / X-Task-Diagnostics feature - dumps task-specific logs into a profile artifact.

docs/dev/learn/tasks/diagnostics.md

@@ -35,7 +35,9 @@ The following diagnostics are supported currently:
 - memray:
    Dumps a profile which can be processed with `memray`, which shows which lines and functions were
    responsible for the most allocations at the time of peak RSS of the process
-
+- logs:
+   Dumps all logs specific to the task, including DEBUG logs, regardless of whether DEBUG logging
+   is otherwise enabled
 
 ## Memory Logging
 

pulpcore/app/settings.py

@@ -374,8 +374,9 @@
 #    lines and functions, at the time of peak RSS of the task process. This adds significant
 #    runtime overhead to the task process, 20-40%. Tweaking code might be warranted for
 #    some advanced settings.
+# * "logs" - Dumps the logs specific to each task. All logs are captured, including DEBUG logs.
 # NOTE: "memray" and "pyinstrument" require additional packages to be installed on the system.
-TASK_DIAGNOSTICS = []  # ["memory", "pyinstrument", "memray"]
+TASK_DIAGNOSTICS = []  # ["memory", "pyinstrument", "memray", "logs"]
 
 ANALYTICS = True
 

pulpcore/tasking/_util.py

@@ -121,6 +121,8 @@ def _execute_task_and_profile(task, profile_options):
             _execute_task = _pyinstrument_diagnostic_decorator(temp_dir, _execute_task)
         if "memray" in profile_options:
             _execute_task = _memray_diagnostic_decorator(temp_dir, _execute_task)
+        if "logs" in profile_options:
+            _execute_task = _logging_decorator(temp_dir, _execute_task)
 
         _execute_task(task)
 
@@ -218,6 +220,50 @@ def __memray_diagnostic_decorator(task):
     return __memray_diagnostic_decorator
 
 
+def _logging_decorator(temp_dir, func):
+    def __logging_decorator(task):
+        log_file_path = os.path.join(temp_dir, "task_logs.log")
+
+        # Create a file handler that captures all logging levels
+        file_handler = logging.FileHandler(log_file_path, mode="w", encoding="utf-8")
+        file_handler.setLevel(logging.NOTSET)  # Capture all levels
+
+        # Create a formatter for consistent log formatting
+        formatter = logging.Formatter(
+            "%(asctime)s - %(name)s - %(levelname)s - %(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+        )
+        file_handler.setFormatter(formatter)
+
+        # Get the root logger to capture all logs
+        root_logger = logging.getLogger()
+        original_level = root_logger.level
+
+        try:
+            # Add the handler to the root logger
+            root_logger.addHandler(file_handler)
+
+            # Execute the task
+            func(task)
+        finally:
+            # Always remove the handler and restore original level
+            root_logger.removeHandler(file_handler)
+            file_handler.close()
+
+        # Save the log file as a ProfileArtifact
+        artifact = Artifact.init_and_validate(log_file_path)
+        try:
+            # it's unlikely for a log file to be identical, but we retain the same check as the
+            # other decorators
+            artifact.save()
+        except IntegrityError:
+            artifact = Artifact.objects.get(sha256=artifact.sha256)
+
+        ProfileArtifact.objects.get_or_create(artifact=artifact, name="task_logs", task=task)
+        _logger.info("Created task logging diagnostic data.")
+
+    return __logging_decorator
+
+
 def dispatch_scheduled_tasks():
     # Warning, dispatch_scheduled_tasks is not race condition free!
     now = timezone.now()