set auto vaccum False, add fail_if_requests_timedout_in_last_x_second…

…s_is_more_than_y

Signed-off-by: Praneeth Bedapudi <[email protected]>

fastdeploy/_loop.py

@@ -132,6 +132,9 @@ def fetch_batch(
             # finished collecting batch
             break
 
+    _utils.logger.info(
+        f"Fetched batch {[v for v in unique_id_wise_input_count.values()]}"
+    )
     return unique_id_wise_input_count, input_batch
 
 
@@ -217,7 +220,7 @@ def start_loop(
             query={
                 "$and": [
                     {"-1.predicted_at": {"$gt": 0}},
-                    {"-1.predicted_at": {"$lt": time.time() - 30}},
+                    {"-1.predicted_at": {"$lt": time.time() - 40}},
                 ]
             },
         )

fastdeploy/_rest.py

@@ -168,7 +168,7 @@ def on_post(self, req, resp):
 class PrometheusMetrics(object):
     def on_get(self, req, resp):
         _LAST_X_SECONDS = int(
-            req.params.get("last_x_seconds", int(os.getenv("LAST_X_SECONDS", 5)))
+            req.params.get("last_x_seconds", int(os.getenv("LAST_X_SECONDS", 30)))
         )
         CURRENT_TIME = time.time()
         LAST_X_SECONDS = time.time() - _LAST_X_SECONDS
@@ -325,6 +325,12 @@ def on_get(self, req, resp):
             "fail_if_up_time_more_than_x_seconds", None
         )
 
+        fail_if_requests_timedout_in_last_x_seconds_is_more_than_y_param = (
+            req.params.get(
+                "fail_if_requests_timedout_in_last_x_seconds_is_more_than_y", None
+            )
+        )
+
         is_predictor_is_up_param = req.params.get("is_predictor_is_up", None)
 
         if fail_if_percentage_of_requests_failed_in_last_x_seconds_is_more_than_y_param:
@@ -342,7 +348,7 @@ def on_get(self, req, resp):
                 resp.media = {
                     "reason": f"More than {y}% requests failed in last {x} seconds"
                 }
-            return
+                return
 
         elif fail_if_requests_older_than_x_seconds_pending_param:
             if _utils.check_if_requests_older_than_x_seconds_pending(
@@ -352,7 +358,7 @@ def on_get(self, req, resp):
                 resp.media = {
                     "reason": f"Requests older than {fail_if_requests_older_than_x_seconds_pending_param} seconds are pending"
                 }
-            return
+                return
 
         elif fail_if_up_time_more_than_x_seconds_param:
             if time.time() - Infer.started_at_time > int(
@@ -362,6 +368,19 @@ def on_get(self, req, resp):
                 resp.media = {
                     "reason": f"Up time more than {fail_if_up_time_more_than_x_seconds_param} seconds"
                 }
+                return
+
+        elif fail_if_requests_timedout_in_last_x_seconds_is_more_than_y_param:
+            (
+                x,
+                y,
+            ) = fail_if_requests_timedout_in_last_x_seconds_is_more_than_y_param.split(
+                ","
+            )
+            x, y = int(x), int(y)
+            if _utils.check_if_requests_timedout_in_last_x_seconds_is_more_than_y(x, y):
+                resp.status = falcon.HTTP_503
+                return
 
         resp.status = falcon.HTTP_200
         resp.media = {"status": "ok"}

fastdeploy/_utils.py

@@ -88,6 +88,7 @@
         **{f"{_}.predicted_in_batch_of": "number" for _ in PREDICTOR_SEQUENCE_TO_FILES},
     },
     db_path=os.path.join("fastdeploy_dbs", f"main_index.db"),
+    auto_vacuum=False,
 )
 
 # for setting timedout_in_queue
@@ -176,6 +177,35 @@ def calculate_optimum_batch_sizes(
     return max_batch_size, time_per_example
 
 
+def check_if_requests_timedout_in_last_x_seconds_is_more_than_y(
+    last_x_seconds, max_percentage_of_timedout_requests
+):
+    time_before_x_seconds = time.time() - last_x_seconds
+    requests_received_in_last_x_seconds = MAIN_INDEX.count(
+        query={"-1.received_at": {"$gte": time_before_x_seconds}}
+    )
+
+    requests_timedout_in_last_x_seconds = MAIN_INDEX.count(
+        query={
+            "-1.received_at": {"$gte": time_before_x_seconds},
+            "timedout_in_queue": True,
+        }
+    )
+
+    if requests_received_in_last_x_seconds == 0:
+        return False
+
+    logger.warning(
+        f"Requests timedout in last {last_x_seconds} seconds: {requests_timedout_in_last_x_seconds}/{requests_received_in_last_x_seconds}"
+    )
+
+    if (
+        requests_timedout_in_last_x_seconds / requests_received_in_last_x_seconds
+    ) * 100 >= max_percentage_of_timedout_requests:
+        return True
+    return False
+
+
 def check_if_percentage_of_requests_failed_in_last_x_seconds_is_more_than_y(
     last_x_seconds, max_percentage_of_failed_requests
 ):

setup.py

@@ -18,10 +18,10 @@
 EMAIL = "[email protected]"
 AUTHOR = "BEDAPUDI PRANEETH"
 REQUIRES_PYTHON = ">=3.6.0"
-VERSION = "3.0.28"
+VERSION = "3.0.30"
 
 # What packages are required for this module to be executed?
-REQUIRED = ["falcon", "liteindex==0.0.3.2.dev4", "zstandard", "gunicorn[gevent]", "msgpack"]
+REQUIRED = ["falcon", "liteindex==0.0.3.2.dev6", "zstandard", "gunicorn[gevent]", "msgpack"]
 
 # What packages are optional?
 EXTRAS = {