Add a beat processing schedule for environment provisioning jobs.

The beat schedule is set to once per minute for each of the three
environment provisioning tasks.

Adding a beat schedule surfaced two problems that are addressed here
with the following changes:
- Commit the SQLALchemy session in order to release the environment
  lock. Otherwise the change to the `claimed_until` field is not
  persisted.
- Set `none_as_null` on the JSOB fields on the `Environment`. This
  avoids problems with querying on Postgres JSON fields that are empty.

This also adds a small change to the development command for the Celery
worker. Multiple child processes were executing the beat jobs, which
lead to exceptions for environment locks and confusing log output. This
contrains the dev command to a single Celery worker.

atst/domain/environments.py

@@ -1,4 +1,4 @@
-from sqlalchemy import text, func, or_
+from sqlalchemy import func, or_
 from sqlalchemy.orm.exc import NoResultFound
 from typing import List
 from uuid import UUID
@@ -130,7 +130,7 @@ class Environments(object):
         results = (
             cls.base_provision_query(now)
             .filter(Environment.cloud_id != None)
-            .filter(Environment.root_user_info == text("'null'"))
+            .filter(Environment.root_user_info == None)
         ).all()
         return [id_ for id_, in results]
 
@@ -143,7 +143,7 @@ class Environments(object):
         results = (
             cls.base_provision_query(now)
             .filter(Environment.cloud_id != None)
-            .filter(Environment.root_user_info != text("'null'"))
-            .filter(Environment.baseline_info == text("'null'"))
+            .filter(Environment.root_user_info != None)
+            .filter(Environment.baseline_info == None)
         ).all()
         return [id_ for id_, in results]

atst/models/environment.py

@@ -26,8 +26,8 @@ class Environment(
     creator = relationship("User")
 
     cloud_id = Column(String)
-    root_user_info = Column(JSONB)
-    baseline_info = Column(JSONB)
+    root_user_info = Column(JSONB(none_as_null=True))
+    baseline_info = Column(JSONB(none_as_null=True))
 
     claimed_until = Column(TIMESTAMP(timezone=True))
 

atst/models/utils.py

@@ -47,3 +47,4 @@ def claim_for_update(resource, minutes=30):
         db.session.query(Model).filter(Model.id == resource.id).filter(
             Model.claimed_until != None
         ).update({"claimed_until": None}, synchronize_session="fetch")
+        db.session.commit()

atst/queue.py

@@ -5,7 +5,20 @@ celery = Celery(__name__)
 
 def update_celery(celery, app):
     celery.conf.update(app.config)
-    celery.conf.CELERYBEAT_SCHEDULE = {}
+    celery.conf.CELERYBEAT_SCHEDULE = {
+        "beat-dispatch_create_environment": {
+            "task": "atst.jobs.dispatch_create_environment",
+            "schedule": 60,
+        },
+        "beat-dispatch_create_atat_admin_user": {
+            "task": "atst.jobs.dispatch_create_atat_admin_user",
+            "schedule": 60,
+        },
+        "beat-dispatch_create_environment_baseline": {
+            "task": "atst.jobs.dispatch_create_environment_baseline",
+            "schedule": 60,
+        },
+    }
 
     class ContextTask(celery.Task):
         def __call__(self, *args, **kwargs):

script/dev_queue

@@ -4,7 +4,7 @@
 
 set -e
 
-WORKER="pipenv run celery -A celery_worker.celery worker --loglevel=info -B"
+WORKER="pipenv run celery -A celery_worker.celery worker --loglevel=info -B -c 1"
 
 if [[ `command -v entr` ]]; then
   find atst | entr -r $WORKER