Switch postgres to a non-transactional DDL to allow us to use peewee to modify data in migrations: enterprise customers are running postgres migrations offline already. Move the image backfill script back to a migration since it will now work. Unify the interface to sending a DB URI to env.py for the migration script.

conf/init/runmigration.sh

@@ -2,7 +2,4 @@
 set -e
 
 # Run the database migration
-PYTHONPATH=. venv/bin/alembic upgrade head
-
-# Run the uncompressed size migration
-PYTHONPATH=. venv/bin/python -m util.uncompressedsize
+PYTHONPATH=. venv/bin/alembic upgrade head

data/migrations/env.py

@@ -13,24 +13,8 @@ from app import app
 from data.model.sqlalchemybridge import gen_sqlalchemy_metadata
 from util.morecollections import AttrDict
 
-# this is the Alembic Config object, which provides
-# access to the values within the .ini file in use.
-db_uri = unquote(app.config['DB_URI'])
-if 'GENMIGRATE' in os.environ:
-    docker_host = os.environ.get('DOCKER_HOST')
-    docker_host_ip = docker_host[len('tcp://'):].split(':')[0]
-    if os.environ.get('GENMIGRATE') == 'mysql':
-        db_uri = 'mysql+pymysql://root:password@%s/genschema' % (docker_host_ip)
-    else:
-        db_uri = 'postgresql://postgres@%s/genschema' % (docker_host_ip)
-
-if 'DB_URI' in os.environ:
-    db_uri = os.environ['DB_URI']
-
-app.config['DB_URI'] = db_uri
-
 config = context.config
-config.set_main_option('sqlalchemy.url', db_uri)
+config.set_main_option('sqlalchemy.url', unquote(app.config['DB_URI']))
 
 # Interpret the config file for Python logging.
 # This line sets up loggers basically.
@@ -86,7 +70,8 @@ def run_migrations_online():
     connection = engine.connect()
     context.configure(
                 connection=connection,
-                target_metadata=target_metadata
+                target_metadata=target_metadata,
+                transactional_ddl=False,
                 )
 
     try:

data/migrations/migration.sh

@@ -1,4 +1,8 @@
-set -e 
+set -e
+
+DOCKER_IP=`echo $DOCKER_HOST | sed 's/tcp:\/\///' | sed 's/:.*//'`
+MYSQL_CONFIG_OVERRIDE="{\"DB_URI\":\"mysql+pymysql://root:password@$DOCKER_IP/genschema\"}"
+PGSQL_CONFIG_OVERRIDE="{\"DB_URI\":\"postgresql://postgres@$DOCKER_IP/genschema\"}"
 
 up_mysql() {
   # Run a SQL database on port 3306 inside of Docker.
@@ -36,19 +40,19 @@ down_postgres() {
 
 gen_migrate() {
   # Generate a SQLite database with the schema as defined by the existing alembic model.
-  GENMIGRATE=$1 PYTHONPATH=. alembic upgrade head
+  QUAY_OVERRIDE_CONFIG=$1 PYTHONPATH=. alembic upgrade head
 
   # Generate the migration to the current model.
-  GENMIGRATE=$1 PYTHONPATH=. alembic revision --autogenerate -m "$2"
+  QUAY_OVERRIDE_CONFIG=$1 PYTHONPATH=. alembic revision --autogenerate -m "$2"
 }
 
 test_migrate() {
   # Generate a SQLite database with the schema as defined by the existing alembic model.
-  GENMIGRATE=$1 PYTHONPATH=. alembic upgrade head
+  QUAY_OVERRIDE_CONFIG=$1 PYTHONPATH=. alembic upgrade head
 
   # Downgrade to verify it works in both directions.
   COUNT=`ls data/migrations/versions/*.py | wc -l | tr -d ' '`
-  GENMIGRATE=$1 PYTHONPATH=. alembic downgrade "-$COUNT"
+  QUAY_OVERRIDE_CONFIG=$1 PYTHONPATH=. alembic downgrade "-$COUNT"
 }
 
 # Test (and generate, if requested) via MySQL.
@@ -59,13 +63,13 @@ if [ ! -z "$@" ]
   then
     set +e
     echo '> Generating Migration'
-    gen_migrate "mysql" "$@"
+    gen_migrate $MYSQL_CONFIG_OVERRIDE "$@"
     set -e
   fi
 
 echo '> Testing Migration (mysql)'
 set +e
-test_migrate "mysql"
+test_migrate $MYSQL_CONFIG_OVERRIDE
 set -e
 down_mysql
 
@@ -75,8 +79,6 @@ up_postgres
 
 echo '> Testing Migration (postgres)'
 set +e
-test_migrate "postgres"
+test_migrate $PGSQL_CONFIG_OVERRIDE
 set -e
 down_postgres
-
-

data/migrations/versions/2430f55c41d5_calculate_uncompressed_sizes_for_all_.py

@@ -16,9 +16,7 @@ from util.uncompressedsize import backfill_sizes_from_data
 
 
 def upgrade(tables):
-  # Note: Doing non-alembic operations inside alembic can cause a deadlock. This call has been
-  # moved to runmigration.sh.
-  pass
+  backfill_sizes_from_data()
 
 def downgrade(tables):
   pass

data/migrations/versions/2fb36d4be80d_translate_the_queue_names_to_reference_.py

@@ -1,14 +1,14 @@
 """Translate the queue names to reference namespace by id, remove the namespace column.
 
 Revision ID: 2430f55c41d5
-Revises: 9a1087b007d
+Revises: 313d297811c4
 Create Date: 2014-09-30 17:31:33.308490
 
 """
 
 # revision identifiers, used by Alembic.
 revision = '2fb36d4be80d'
-down_revision = '2430f55c41d5'
+down_revision = '313d297811c4'
 
 from alembic import op
 import sqlalchemy as sa

util/uncompressedsize.py

@@ -1,9 +1,8 @@
 import logging
 import zlib
-import sys
 
 from data import model
-from data.database import ImageStorage, configure
+from data.database import ImageStorage
 from app import app, storage as store
 from data.database import db, db_random_func
 from util.gzipstream import ZLIB_GZIP_WINDOW
@@ -15,16 +14,11 @@ logger = logging.getLogger(__name__)
 CHUNK_SIZE = 5 * 1024 * 1024
 
 def backfill_sizes_from_data():
-  logger.setLevel(logging.DEBUG)  
+  logger.setLevel(logging.DEBUG)
 
   logger.debug('Starting uncompressed image size backfill')
   logger.debug('NOTE: This can be a LONG RUNNING OPERATION. Please wait!')
 
-  # Make sure we have a reference to the current DB.
-  configure(app.config)
-
-  logger.debug('Uncompressed backfill: Database configured')
-
   # Check for any uncompressed images.
   has_images = bool(list(ImageStorage
                          .select(ImageStorage.uuid)