fix db connection pool

app/__init__.py

@@ -18,6 +18,7 @@ from sqlalchemy import event
 from werkzeug.exceptions import HTTPException as WerkzeugHTTPException
 from werkzeug.local import LocalProxy
 
+from app import config
 from app.clients import NotificationProviderClients
 from app.clients.cloudwatch.aws_cloudwatch import AwsCloudwatchClient
 from app.clients.document_download import DocumentDownloadClient
@@ -25,6 +26,7 @@ from app.clients.email.aws_ses import AwsSesClient
 from app.clients.email.aws_ses_stub import AwsSesStubClient
 from app.clients.pinpoint.aws_pinpoint import AwsPinpointClient
 from app.clients.sms.aws_sns import AwsSnsClient
+from app.utils import hilite
 from notifications_utils import logging, request_helper
 from notifications_utils.clients.encryption.encryption_client import Encryption
 from notifications_utils.clients.redis.redis_client import RedisClient
@@ -58,15 +60,29 @@ class SQLAlchemy(_SQLAlchemy):
 
     def apply_driver_hacks(self, app, info, options):
         sa_url, options = super().apply_driver_hacks(app, info, options)
+        print(hilite(f"OPTIONS {options}"))
+
         if "connect_args" not in options:
             options["connect_args"] = {}
         options["connect_args"]["options"] = "-c statement_timeout={}".format(
             int(app.config["SQLALCHEMY_STATEMENT_TIMEOUT"]) * 1000
         )
+
         return (sa_url, options)
 
 
-db = SQLAlchemy()
+# Set db engine settings here for now.
+# They were not being set previous (despite environmental variables with appropriate
+# sounding names) and were defaulting to low values
+db = SQLAlchemy(
+    engine_options={
+        "pool_size": config.Config.SQLALCHEMY_POOL_SIZE,
+        "max_overflow": 10,
+        "pool_timeout": config.Config.SQLALCHEMY_POOL_TIMEOUT,
+        "pool_recycle": config.Config.SQLALCHEMY_POOL_RECYCLE,
+        "pool_pre_ping": True,
+    }
+)
 migrate = Migrate()
 ma = Marshmallow()
 notify_celery = NotifyCelery()

app/celery/scheduled_tasks.py

@@ -243,6 +243,8 @@ def check_for_services_with_high_failure_rates_or_sending_to_tv_numbers():
     bind=True, max_retries=7, default_retry_delay=3600, name="process-delivery-receipts"
 )
 def process_delivery_receipts(self):
+    # If we need to check db settings do it here for convenience
+    # current_app.logger.info(f"POOL SIZE {app.db.engine.pool.size()}")
     """
     Every eight minutes or so (see config.py) we run this task, which searches the last ten
     minutes of logs for delivery receipts and batch updates the db with the results.  The overlap

app/clients/__init__.py

@@ -13,8 +13,7 @@ AWS_CLIENT_CONFIG = Config(
         "addressing_style": "virtual",
     },
     use_fips_endpoint=True,
-    # This is the default but just for doc sake
+    max_pool_connections=50,  # This should be equal or greater than our celery concurrency
-    max_pool_connections=10,
 )
 
 

app/config.py

@@ -81,7 +81,7 @@ class Config(object):
     SQLALCHEMY_DATABASE_URI = cloud_config.database_url
     SQLALCHEMY_RECORD_QUERIES = False
     SQLALCHEMY_TRACK_MODIFICATIONS = False
-    SQLALCHEMY_POOL_SIZE = int(getenv("SQLALCHEMY_POOL_SIZE", 5))
+    SQLALCHEMY_POOL_SIZE = int(getenv("SQLALCHEMY_POOL_SIZE", 20))
     SQLALCHEMY_POOL_TIMEOUT = 30
     SQLALCHEMY_POOL_RECYCLE = 300
     SQLALCHEMY_STATEMENT_TIMEOUT = 1200