Merge pull request #1862 from GSA/try_gevent

try gevent

Author: ccostino

Makefile

@@ -70,7 +70,7 @@ run-celery: ## Run celery, TODO remove purge for staging/prod
 		-A run_celery.notify_celery worker \
 		--pidfile="/tmp/celery.pid" \
 		--loglevel=INFO \
-		--pool=eventlet
+		--pool=gevent
 		--concurrency=20
 
 

app/aws/s3.py

@@ -5,7 +5,7 @@
 from io import StringIO
 
 import botocore
-import eventlet
+import gevent
 from boto3 import Session
 from flask import current_app
 
@@ -250,7 +250,7 @@ def get_s3_files():
         for object_key in object_keys:
             read_s3_file(bucket_name, object_key, s3res)
             count = count + 1
-            eventlet.sleep(0.2)
+            gevent.sleep(0.2)
     except Exception:
         current_app.logger.exception(
             f"Trouble reading {object_key} which is # {count} during cache regeneration"
@@ -411,7 +411,7 @@ def get_job_from_s3(service_id, job_id):
                 )
                 retries += 1
                 sleep_time = backoff_factor * (2**retries)  # Exponential backoff
-                eventlet.sleep(sleep_time)
+                gevent.sleep(sleep_time)
                 continue
             else:
                 # Typically this is "NoSuchKey"

app/celery/tasks.py

@@ -1,6 +1,6 @@
 import json
 
-import eventlet
+import gevent
 from celery.signals import task_postrun
 from flask import current_app
 from requests import HTTPError, RequestException, request
@@ -84,7 +84,7 @@ def process_job(job_id, sender_id=None):
         process_row(row, template, job, service, sender_id=sender_id)
         count = count + 1
         if count % 3 == 0:
-            eventlet.sleep(1)
+            gevent.sleep(1)
 
     # End point/Exit point for message send flow.
     job_complete(job, start=start)

app/clients/cloudwatch/aws_cloudwatch.py

@@ -151,7 +151,7 @@ def event_to_db_format(self, event):
     #         result = temp_client.get_query_results(queryId=query_id)
     #         if result['status'] == 'Complete':
     #             break
-    #         eventlet.sleep(1)
+    #         gevent.sleep(1)
 
     #     delivery_receipts = []
     #     for log in result['results']:

gunicorn_config.py

@@ -1,19 +1,17 @@
-import eventlet
+from gevent import monkey
 
-eventlet.monkey_patch()  # this has to be called before other imports or monkey patching doesn't happen
+monkey.patch_all()  # this has to be called before other imports or monkey patching doesn't happen
 
 import os  # noqa
 import socket  # noqa
 import sys  # noqa
 import traceback  # noqa
 
-import eventlet.debug as debug  # noqa
 import gunicorn  # noqa
 
-# This will give us a better stack trace if blocking occurs (if True)
-debug.hub_blocking_detection(False)
+# This will give us a better stack trace if
 workers = 4
-worker_class = "eventlet"
+worker_class = "gevent"
 worker_connections = 256
 bind = "0.0.0.0:{}".format(os.getenv("PORT"))
 statsd_host = "{}:8125".format(os.getenv("STATSD_HOST"))
@@ -36,22 +34,3 @@ def on_exit(server):
 
 def worker_int(worker):
     worker.log.info("worker: received SIGINT {}".format(worker.pid))
-
-
-def fix_ssl_monkeypatching():
-    """
-    eventlet works by monkey-patching core IO libraries (such as ssl) to be non-blocking. However, there's currently
-    a bug: In the normal socket library it may throw a timeout error as a `socket.timeout` exception. However
-    eventlet.green.ssl's patch raises an ssl.SSLError('timed out',) instead. redispy handles socket.timeout but not
-    ssl.SSLError, so we solve this by monkey patching the monkey patching code to raise the correct exception type
-    :scream:
-
-    https://github.com/eventlet/eventlet/issues/692
-    """
-    # this has probably already been called somewhere in gunicorn internals, however, to be sure, we invoke it again.
-    # eventlet.monkey_patch can be called multiple times without issue
-    eventlet.monkey_patch()
-    eventlet.green.ssl.timeout_exc = socket.timeout
-
-
-fix_ssl_monkeypatching()

manifest.yml

@@ -26,7 +26,7 @@ applications:
       - type: worker
         instances: ((worker_instances))
         memory: ((worker_memory))
-        command: newrelic-admin run-program celery -A run_celery.notify_celery worker --loglevel=INFO --pool=eventlet --concurrency=20 --prefetch-multiplier=2
+        command: newrelic-admin run-program celery -A run_celery.notify_celery worker --loglevel=INFO --pool=gevent --concurrency=20 --prefetch-multiplier=2
       - type: scheduler
         instances: 1
         memory: ((scheduler_memory))

poetry.lock

@@ -25,14 +25,14 @@ click-didyoumean = "==0.3.1"
 click-plugins = "==1.1.1.2"
 click-repl = "==0.3.0"
 deprecated = "==1.2.18"
-eventlet = "==0.40.2"
+gevent = "==25.5.1"
 expiringdict = "==1.2.2"
 flask = "~=3.1"
 flask-bcrypt = "==1.0.1"
 flask-migrate = "==4.1.0"
 flask-redis = "==0.4.0"
 flask-sqlalchemy = "^3.1.1"
-gunicorn = {version = "==23.0.0", extras = ["eventlet"]}
+gunicorn = {version = "==23.0.0", extras = ["gevent"]}
 iso8601 = "==2.1.0"
 jsonschema = {version = "==4.25.0", extras = ["format"]}
 lxml = "==6.0.0"