amqp/adapter/backpressure_handler.py

@@ -1,23 +1,17 @@
 import asyncio
 import json
-import logging
-import os
 import time
 from asyncio import AbstractEventLoop
 from threading import Thread
 
 from aio_pika import Message
-from aio_pika.abc import AbstractRobustChannel, ExchangeType
+from aio_pika.abc import AbstractRobustChannel
 
 from amqp.adapter.logging_utils import logging_info, logging_warning
+from amqp.config.amq_configuration import AMQConfiguration
 from amqp.model.model import ScalingRequestAlert
 from amqp.router.utils import await_future, await_result
 
-# Duration in seconds between backpressure reports
-REPORT_WINDOW_SEC = 10
-# Number of reports without activity to trigger idle event
-NUM_REPORTS_FOR_IDLE = 3
-
 
 class ScaleRequestV1:
 
@@ -61,15 +55,20 @@ class BackpressureHandler:
     last_backpressure_event_time = 0
     last_backpressure_event = ScalingRequestAlert.UPDATE
 
-    parallel_workers = int(os.getenv("PARALLEL_WORKERS", default=0))
-    swarm_service_id = os.getenv("SWARM_SERVICE_ID", default="clever-amqp-service")
-    swarm_task_id = os.getenv("SWARM_TASK_ID", default="python-adapter")
-
-    def __init__(self, channel: AbstractRobustChannel, loop: AbstractEventLoop):
+    def __init__(
+        self,
+        channel: AbstractRobustChannel,
+        loop: AbstractEventLoop,
+        config: AMQConfiguration,
+    ):
         self.lock = None
         self.channel = channel
         self.loop = loop
+        self.config = config
         self.exchange = None
+        self.swarm_service_id = self.config.amq_adapter.swarm_service_id
+        self.swarm_task_id = self.config.amq_adapter.swarm_task_id
+        self.parallel_workers = self.config.backpressure.threshold_threads
 
     def increase_parallel_executions(self):
         """Increase the number of parallel executions"""
@@ -120,7 +119,8 @@ class BackpressureHandler:
                 self.last_backpressure_event = ScalingRequestAlert.OVERLOAD
 
     def backpressure_monitor_loop(self):
-        global REPORT_WINDOW_SEC
+        _time_window_sec = self.config.backpressure.time_window
+        _idle_duration_millis = 1000 * self.config.backpressure.idle_duration
         _loop = asyncio.new_event_loop()
 
         async def _backpressure_monitor():
@@ -143,7 +143,7 @@ class BackpressureHandler:
                 if (
                     self.current_parallel_executions >= self.parallel_workers
                     and time.time() - self.last_backpressure_event_time
-                    > REPORT_WINDOW_SEC
+                    > _time_window_sec
                     and self.last_backpressure_event != ScalingRequestAlert.OVERLOAD
                 ):
                     # Trigger the overload event
@@ -153,7 +153,7 @@ class BackpressureHandler:
                 elif (
                     self.current_parallel_executions == 0
                     and time.time() - self.last_data_message_time
-                    > NUM_REPORTS_FOR_IDLE * REPORT_WINDOW_SEC
+                    >= _idle_duration_millis
                     and self.last_backpressure_event != ScalingRequestAlert.IDLE
                 ):
                     # Trigger the idle event
@@ -165,7 +165,7 @@ class BackpressureHandler:
                     if (
                         self.current_parallel_executions > 0
                         and time.time() - self.last_backpressure_event_time
-                        > REPORT_WINDOW_SEC
+                        > _time_window_sec
                     ):
                         # Trigger the update event
                         _scaling_request: ScaleRequestV1 = ScaleRequestV1(
@@ -179,7 +179,7 @@ class BackpressureHandler:
                         await self.publish_backpressure_request(_scaling_request)
                         self.last_backpressure_event_time = time.time()
                         self.last_backpressure_event = ScalingRequestAlert.UPDATE
-                await asyncio.sleep(REPORT_WINDOW_SEC)
+                await asyncio.sleep(_time_window_sec)
 
         _loop.run_until_complete(_backpressure_monitor())
 

amqp/adapter/cleverthis_service_adapter.py

@@ -6,7 +6,6 @@ Methods here will invoke the actual CleverThis Service code, which are passed as
 import asyncio
 import inspect
 import json
-import logging
 from asyncio import AbstractEventLoop
 from typing import Dict, List, Callable, Coroutine, Any
 
@@ -141,23 +140,9 @@ class CleverThisServiceAdapter:
         self.service_host = self.amq_configuration.dispatch.service_host
         self.service_port = self.amq_configuration.dispatch.service_port
         self.loop: AbstractEventLoop | None = None
-        try:
-            self.require_authenticated_user = (
-                self.amq_configuration.amq_adapter.require_authenticated_user
-                and isinstance(
-                    self.amq_configuration.amq_adapter.require_authenticated_user, bool
-                )
-                or eval(
-                    str(self.amq_configuration.amq_adapter.require_authenticated_user)
-                    .strip()
-                    .capitalize()
-                    or "True"
-                )
-            )
-        except Exception as e:
-            self.require_authenticated_user = (
-                True  # Enforce authenticated user as default
-            )
+        self.require_authenticated_user = (
+            self.amq_configuration.amq_adapter.require_authenticated_user
+        )
 
     def get_content_type(self, message_headers: Dict[str, List[str]]) -> str:
         """

amqp/config/amq_configuration.py

@@ -1,6 +1,5 @@
 import configparser
 import inspect
-import logging
 import os
 
 from amqp.adapter.logging_utils import logging_error, logging_warning
@@ -36,11 +35,16 @@ class AMQAdapter:
         self.route_mapping = config.get(
             "CleverMicro-AMQ", self.PREFIX + "route-mapping", fallback=None
         )
-        self.require_authenticated_user = config.get(
+        self.require_authenticated_user = config.getboolean(
             "CleverMicro-AMQ",
             self.PREFIX + "require-authenticated-user",
             fallback=False,
         )
+        # these 2 shall be provided by Docker swarm, format is a random alphanumeric string
+        self.swarm_service_id = os.getenv(
+            "SWARM_SERVICE_ID", default="clever-amqp-service"
+        )
+        self.swarm_task_id = os.getenv("SWARM_TASK_ID", default="python-adapter")
 
     def adapter_prefix(self) -> str:
         return f"{self.service_name}-{self.generator_id}-"
@@ -56,7 +60,6 @@ class Dispatch:
     def __init__(self, config: configparser.ConfigParser):
         """
         cm.dispatch.use-dlq=true
-        cm.dispatch.use-confirms=false
         cm.dispatch.amq-host=rabbitmq
         cm.dispatch.amq-port=5672
         cm.dispatch.amq-port-tls=5671
@@ -68,9 +71,6 @@ class Dispatch:
         self.use_dlq = config.getboolean(
             "CleverMicro-AMQ", self.PREFIX + "use-dlq", fallback=False
         )
-        self.use_confirms = config.getboolean(
-            "CleverMicro-AMQ", self.PREFIX + "use-confirms", fallback=False
-        )
         self.amq_host = config.get(
             "CleverMicro-AMQ", self.PREFIX + "amq-host", fallback="rabbitmq"
         )
@@ -110,12 +110,27 @@ class Backpressure:
 
         :param config: config parser to read configuration values
         """
-        self.threshold = config.getint(
-            "CleverMicro-AMQ", self.PREFIX + "threshold", fallback=0
+        self.threshold_threads = config.getint(
+            "CleverMicro-AMQ",
+            self.PREFIX + "threshold",
+            fallback=int(os.getenv("PARALLEL_WORKERS", 1)),
         )
-        # time-window is in milliseconds, to report backpressure IDLE alert
+        # time-window in seconds, duration between backpressure reports
         self.time_window = config.getint(
-            "CleverMicro-AMQ", self.PREFIX + "time-window", fallback=10000
+            "CleverMicro-AMQ", self.PREFIX + "time-window", fallback=10
+        )
+
+        # Define maximum CPU usage before triggering backpressure OVERLOAD alert
+        self.threshold_cpu = config.getint(
+            "CleverMicro-AMQ", self.PREFIX + "threshold-cpu", fallback=90
+        )
+        # Define the time window length, in seconds, during which, if CPU usage is consistently above the threshold, the system will trigger OVERLOAD alert
+        self.cpu_overload_duration = config.getint(
+            "CleverMicro-AMQ", self.PREFIX + "cpu-overload-duration", fallback=30
+        )
+        # Define the time window length, in seconds, during which, if no request is made to the Service, the system will trigger IDLE alert
+        self.idle_duration = config.getint(
+            "CleverMicro-AMQ", self.PREFIX + "idle-duration", fallback=30
         )
 
 

amqp/config/application.properties

@@ -19,4 +19,13 @@ cm.dispatch.download-dir=/tmp/downloaded_files
 #cm.dispatch.service-port=8080
 #
 # Define maximum number of messages of being processed in parallel before triggering backpressure OVERLOAD alert
-cm.backpressure.threshold=20
+cm.backpressure.threshold=5
+# Define maximum CPU usage before triggering backpressure OVERLOAD alert
+cm.backpressure.threshold-cpu=90
+# Define the time window between the Backpressure reports, in seconds
+cm.backpressure.time-window=10
+# Define the time window length, in seconds, during which, if CPU usage is consistently above the threshold, the system will trigger OVERLOAD alert
+cm.backpressure.cpu-overload-duration=30
+# Define the time window length, in seconds, during which, if no request is made to the Service, the system will trigger IDLE alert
+cm.backpressure.idle-duration=30
+

amqp/service/amq_message_handler.py

@@ -1,8 +1,6 @@
 import asyncio
-import logging
 import threading
 import time
-import concurrent.futures
 import os
 from asyncio import Future, AbstractEventLoop
 from typing import Dict
@@ -30,11 +28,6 @@ from amqp.model.model import (
 from amqp.rabbitmq.rabbit_mq_client import RabbitMQClient
 from amqp.adapter.serializer import map_as_string
 
-# Shared thread pool for parallel execution
-executor = concurrent.futures.ThreadPoolExecutor(
-    max_workers=BackpressureHandler.parallel_workers + 2
-)
-
 
 def collect_trace_info():
     trace_map = {}
@@ -79,20 +72,8 @@ class DataMessageHandler:
         # Dictionary to store outstanding Futures
         self.outstanding: Dict[str, asyncio.Future] = {}
         self.reply_to: str | None = None
-        self.backpressure_monitor_window = (
-            amq_configuration.backpressure.time_window / 1000
-        )  # convert to seconds
         self.file_handler: FileHandler = file_handler
         self.backpressure_handler: BackpressureHandler = backpressure_handler
-        _backpressure_treshold = amq_configuration.backpressure.threshold
-        """
-        Determine the number of parallel workers based on the backpressure threshold and PARALLEL_WORKERS env var.
-        If env var PARALLEL_WORKERS is not set (or set to 0), use the backpressure threshold config
-        as the number of parallel workers.
-        """
-        if _backpressure_treshold > 0:
-            if BackpressureHandler.parallel_workers <= 0:
-                BackpressureHandler.parallel_workers = _backpressure_treshold
 
     async def inbound_data_message_callback(self, message: AbstractIncomingMessage):
         threading.Thread(

amqp/service/amq_service.py

@@ -70,7 +70,9 @@ class AMQService:
             loop
         )
         backpressure_handler = BackpressureHandler(
-            channel=self.rabbit_mq_client.channel, loop=loop
+            channel=self.rabbit_mq_client.channel,
+            loop=loop,
+            config=self.amq_configuration,
         )
         self.amq_data_message_handler = DataMessageHandler(
             service_adapter,

pyproject.toml

@@ -8,7 +8,7 @@ packages = ["amqp","amqp.adapter","amqp.config","amqp.model","amqp.rabbitmq","am
 
 [project]
 name = "amq_adapter"
-version = "0.2.22"
+version = "0.2.23"
 description = "The Python implementation of the AMQ Adaptor for CleverMicro"
 readme = "README.md"
 license = { text = "Apache License 2.0" }