Do django db_index migrations run concurrently? - django

Question

I'm looking to add a multi-column index to a postgres database. I have a non blocking SQL command to do this which looks like this:
CREATE INDEX CONCURRENTLY shop_product_fields_index ON shop_product (id, ...);
When I add db_index to my model and run the migration, will it also run concurrently or will it block writes? Is a concurrent migration possible in django?

Answer 1

There are AddIndexConcurrently and RemoveIndexConcurrently in Django 3.0:
https://docs.djangoproject.com/en/dev/ref/contrib/postgres/operations/#django.contrib.postgres.operations.AddIndexConcurrently
Create a migration and then change migrations.AddIndex to AddIndexConcurrently. Import it from django.contrib.postgres.operations.

Answer 2

With Django 1.10 migrations you can create a concurrent index by using RunSQL and disabling the wrapping transaction by making the migration non-atomic by setting atomic = False as a data attribute on the migration:
class Migration(migrations.Migration):
atomic = False # disable transaction
dependencies = []
operations = [
migrations.RunSQL('CREATE INDEX CONCURRENTLY ...')
]
RunSQL: https://docs.djangoproject.com/en/stable/ref/migration-operations/#runsql
Non-atomic Migrations: https://docs.djangoproject.com/en/stable/howto/writing-migrations/#non-atomic-migrations

Answer 3

You could use the SeparateDatabaseAndState migration operation to provide a custom SQL command for creating the index. The operation accepts two lists of operations:
state_operations are operations to apply on the Django model state.
They do not affect the database.
database_operations are operations to apply to the database.
An example migration may look like this:
from django.db import migrations, models
class Migration(migrations.Migration):
atomic = False
dependencies = [
('myapp', '0001_initial'),
]
operations = [
migrations.SeparateDatabaseAndState(
state_operations=[
# operation generated by `makemigrations` to create an ordinary index
migrations.AlterField(
# ...
),
],
database_operations=[
# operation to run custom SQL command (check the output of `sqlmigrate`
# to see the auto-generated SQL, edit as needed)
migrations.RunSQL(sql='CREATE INDEX CONCURRENTLY ...',
reverse_sql='DROP INDEX ...'),
],
),
]

Answer 4

Do what tgroshon says for new django 1.10 +
for lesser versions of django i have had success with a more verbose subclassing method:
from django.db import migrations, models
class RunNonAtomicSQL(migrations.RunSQL):
def _run_sql(self, schema_editor, sqls):
if schema_editor.connection.in_atomic_block:
schema_editor.atomic.__exit__(None, None, None)
super(RunNonAtomicSQL, self)._run_sql(schema_editor, sqls)
class Migration(migrations.Migration):
dependencies = [
]
operations = [
RunNonAtomicSQL(
"CREATE INDEX CONCURRENTLY",
)
]

Answer 5

You can do something like
import django.contrib.postgres.indexes
from django.db import migrations, models
from django.contrib.postgres.operations import AddIndexConcurrently
class Migration(migrations.Migration):
atomic = False
dependencies = [
("app_name", "parent_migration"),
]
operations = [
AddIndexConcurrently(
model_name="mymodel",
index=django.contrib.postgres.indexes.GinIndex(
fields=["field1"],
name="field1_idx",
),
),
AddIndexConcurrently(
model_name="mymodel",
index=models.Index(
fields=["field2"], name="field2_idx"
),
),
]
Ref: https://docs.djangoproject.com/en/dev/ref/contrib/postgres/operations/#django.contrib.postgres.operations.AddIndexConcurrently

Answer 6

There is no support for PostgreSQL concurent index creation in django.
Here is the ticket requesting this feature - https://code.djangoproject.com/ticket/21039
But instead, you can manually specify any custom RunSQL operation in the migration -
https://docs.djangoproject.com/en/stable/ref/migration-operations/#runsql