config: improve gunicorn configs

configs/gunicorn_config.py

@@ -77,14 +77,14 b' worker_connections = 10'
 # Max number of requests that worker will handle before being gracefully restarted.
 # Prevents memory leaks, jitter adds variability so not all workers are restarted at once.
 max_requests = 2000
-max_requests_jitter = 30
+max_requests_jitter = int(max_requests * 0.2)  # 20% of max_requests
 
 # The maximum number of pending connections.
 # Exceeding this number results in the client getting an error when attempting to connect.
 backlog = 64
 
-# Amount of time a worker can spend with handling a request before it
-# gets killed and restarted. By default set to 21600 (6hrs)
+# The Amount of time a worker can spend with handling a request before it
+# gets killed and restarted. By default, set to 21600 (6hrs)
 # Examples: 1800 (30min), 3600 (1hr), 7200 (2hr), 43200 (12h)
 timeout = 21600
 
@@ -124,7 +124,7 b' memory_usage_check_interval = 60'
 
 # Threshold value for which we don't recycle worker if GarbageCollection
 # frees up enough resources. Before each restart, we try to run GC on worker
-# in case we get enough free memory after that, restart will not happen.
+# in case we get enough free memory after that; restart will not happen.
 memory_usage_recovery_threshold = 0.8
 
 
@@ -322,7 +322,7 b' def _check_memory_usage(worker):'
 def worker_int(worker):
     worker.log.info("pid=[%-10s] worker received INT or QUIT signal", worker.pid)
 
-    # get traceback info, on worker crash
+    # get traceback info, when a worker crashes
     def get_thread_id(t_id):
         id2name = dict([(th.ident, th.name) for th in threading.enumerate()])
         return id2name.get(t_id, "unknown_thread_id")