feat: multiprocessing for RabbitMQ worker

Helps with very long tasks

optimizer_worker.py

@@ -2,6 +2,8 @@ from tempfile import TemporaryDirectory
 import json
 import time
 import logging
+import multiprocessing
+import queue
 
 from src.git import clone_repo
 from src.optipng import apply_optipng_recursive
@@ -9,6 +11,17 @@ from src.messages import AnalyzeTask, AnalyzeResponse, create_analyze_response
 from src.rabbit import get_channel, send_reply
 from src.settings import RABBIT_TASK_QUEUE, init_logging, WORKER_NAME
 
+TASK_DELAY = 0.25
+
+
+def delay_generator():
+    for _ in range(10):
+        yield 0.05
+    for _ in range(10):
+        yield 0.1
+    while True:
+        yield TASK_DELAY
+
 
 def process_task(task: AnalyzeTask) -> AnalyzeResponse:
     with TemporaryDirectory() as repo_path:
@@ -27,9 +40,7 @@ def process_task(task: AnalyzeTask) -> AnalyzeResponse:
         )
 
 
-def rabbit_callback(ch, method, properties, body):
+def _process_main(results_queue: queue.Queue, task: AnalyzeTask):
-    task = AnalyzeTask.from_data(json.loads(body))
-    logging.info(f"New task url: {task.url}, level: {task.level}")
     try:
         res = process_task(task).data
         res["status"] = "success"
@@ -40,6 +51,23 @@ def rabbit_callback(ch, method, properties, body):
             "message": str(err),
             "task_id": task.task_id
         }
+    results_queue.put(res)
+
+
+def rabbit_callback(ch, method, properties, body):
+    task = AnalyzeTask.from_data(json.loads(body))
+    logging.info(f"New task url: {task.url}, level: {task.level}")
+    res_queue = multiprocessing.Queue()
+    method_process = multiprocessing.Process(
+        target=_process_main,
+        args=(res_queue, task)
+    )
+    method_process.start()
+    for delay in delay_generator():
+        ch.connection.sleep(delay)
+        if not method_process.is_alive():
+            break
+    res = res_queue.get()
     send_reply(ch, json.dumps(res).encode("utf-8"))
     ch.basic_ack(delivery_tag=method.delivery_tag)