NOT NULL constraints

All attributes that should not have NULL as a value, should be defined as NOT NULL columns in the database.

Depending on the application logic, NOT NULL columns should either have a presence: true validation defined in their Model or have a default value as part of their database definition. As an example, the latter can be true for boolean attributes that should always have a non-NULL value, but have a well defined default value that the application does not need to enforce each time (for example, active=true).

Create a new table with NOT NULL columns

When adding a new table, all NOT NULL columns should be defined as such directly inside create_table.

For example, consider a migration that creates a table with two NOT NULL columns, db/migrate/20200401000001_create_db_guides.rb:

class CreateDbGuides < Gitlab::Database::Migration[2.1]
  def change
    create_table :db_guides do |t|
      t.bigint :stars, default: 0, null: false
      t.bigint :guide, null: false
    end
  end
end

Add a NOT NULL column to an existing table

With PostgreSQL 11 being the minimum version in GitLab, adding columns with NULL and/or default values has become much easier and the standard add_column helper should be used in all cases.

For example, consider a migration that adds a new NOT NULL column active to table db_guides, db/migrate/20200501000001_add_active_to_db_guides.rb:

class AddExtendedTitleToSprints < Gitlab::Database::Migration[2.1]
  def change
    add_column :db_guides, :active, :boolean, default: true, null: false
  end
end

Add a NOT NULL constraint to an existing column

Adding NOT NULL to existing database columns usually requires multiple steps split into at least two different releases. If your table is small enough that you don’t need to use a background migration, you can include all these in the same merge request. We recommend to use separate migrations to reduce transaction durations.

The steps required are:

  1. Release N.M (current release)

    1. Ensure $ATTRIBUTE value is being set at the application level.
      1. If the attribute has a default value, add the default value to the model so the default value is set for new records.
      2. Update all places in the code where the attribute would be set to nil, if any, for new and existing records. Note that using ActiveRecord callbacks such as before_save and before_validation may not be sufficient, as some processes skip these callbacks. update_column, update_columns, and bulk operations such as insert_all and update_all are some examples of methods to look out for.
    2. Add a post-deployment migration to fix the existing records.
    note
    Depending on the size of the table, a background migration for cleanup could be required in the next release. See the NOT NULL constraints on large tables section for more information.
  2. Release N.M+1 (next release)

    1. Make sure all existing records on GitLab.com have attribute set. If not, go back to step 1 from Release N.M.
    2. If step 1 seems fine and the backfill from Release N.M was done via a batched background migration then add a post-deployment migration to finalize the background migration.
    3. Add a validation for the attribute in the model to prevent records with nil attribute as now all existing and new records should be valid.
    4. Add a post-deployment migration to add the NOT NULL constraint.

Example

Considering a given release milestone, such as 13.0.

After checking our production database, we know that there are epics with NULL descriptions, so we cannot add and validate the constraint in one step.

note
Even if we did not have any epic with a NULL description, another instance of GitLab could have such records, so we would follow the same process either way.

Prevent new invalid records (current release)

Update all the code paths where the attribute is being set to nil, if any, to set the attribute to non-nil value for new and existing records.

An attribute with default using the Rails attributes API has been added in epic.rb so that default value is set for new records:

class Epic < ApplicationRecord
  attribute :description, default: 'No description'
end

Data migration to fix existing records (current release)

The approach here depends on the data volume and the cleanup strategy. The number of records that must be fixed on GitLab.com is a nice indicator that helps us decide whether to use a post-deployment migration or a background data migration:

  • If the data volume is less than 1000 records, then the data migration can be executed within the post-migration.
  • If the data volume is higher than 1000 records, it’s advised to create a background migration.

When unsure about which option to use, contact the Database team for advice.

Back to our example, the epics table is not considerably large nor frequently accessed, so we add a post-deployment migration for the 13.0 milestone (current), db/post_migrate/20200501000002_cleanup_epics_with_null_description.rb:

class CleanupEpicsWithNullDescription < Gitlab::Database::Migration[2.1]
  # With BATCH_SIZE=1000 and epics.count=29500 on GitLab.com
  # - 30 iterations will be run
  # - each requires on average ~150ms
  # Expected total run time: ~5 seconds
  BATCH_SIZE = 1000

  disable_ddl_transaction!

  class Epic < ActiveRecord::Base
    include EachBatch

    self.table_name = 'epics'
  end

  def up
    Epic.each_batch(of: BATCH_SIZE) do |relation|
      relation.
        where('description IS NULL').
        update_all(description: 'No description')
    end
  end

  def down
    # no-op : can't go back to `NULL` without first dropping the `NOT NULL` constraint
  end
end

Check if all records are fixed (next release)

Use postgres.ai to create a thin clone of the production database and check if all records on GitLab.com have the attribute set. If not go back to Prevent new invalid records step and figure out where in the code the attribute is explicitly set to nil. Fix the code path then reschedule the migration to fix the existing records and wait for the next release to do the following steps.

Finalize the background migration (next release)

If the migration was done using a background migration then finalize the migration.

Add validation to the model (next release)

Add a validation for the attribute to the model to prevent records with nil attribute as now all existing and new records should be valid.

class Epic < ApplicationRecord
  validates :description, presence: true
end

Add the NOT NULL constraint (next release)

Adding the NOT NULL constraint scans the whole table and make sure that each record is correct.

Still in our example, for the 13.1 milestone (next), we run the add_not_null_constraint migration helper in a final post-deployment migration:

class AddNotNullConstraintToEpicsDescription < Gitlab::Database::Migration[2.1]
  disable_ddl_transaction!

  def up
    # This will add the `NOT NULL` constraint and validate it
    add_not_null_constraint :epics, :description
  end

  def down
    # Down is required as `add_not_null_constraint` is not reversible
    remove_not_null_constraint :epics, :description
  end
end

NOT NULL constraints on large tables

If you have to clean up a nullable column for a high-traffic table (for example, the artifacts in ci_builds), your background migration goes on for a while and it needs an additional batched background migration cleaning up in the release after adding the data migration.

In this case the number of releases depends on the amount of time needed to migrate existing records. The cleanup is scheduled after the background migration has completed, which could be several releases after the constraint was added.

  1. Release N.M:
    • Add the background-migration to fix the existing records:

      # db/post_migrate/
      class QueueBackfillMergeRequestDiffsProjectId < Gitlab::Database::Migration[2.2]
        milestone '16.7'
        restrict_gitlab_migration gitlab_schema: :gitlab_main
      
        MIGRATION = 'BackfillMergeRequestDiffsProjectId'
        DELAY_INTERVAL = 2.minutes
      
        def up
          queue_batched_background_migration(
            MIGRATION,
            :merge_request_diffs,
            :id,
            job_interval: DELAY_INTERVAL
          )
        end
      
        def down
          delete_batched_background_migration(MIGRATION, :merge_request_diffs, :id, [])
        end
      end
      
  2. Release N.M+X, where X is the number of releases the migration was running:
    • Verify that all existing records are fixed.

    • Cleanup the background migration:

      # db/post_migrate/
      class FinalizeMergeRequestDiffsProjectIdBackfill < Gitlab::Database::Migration[2.2]
        disable_ddl_transaction!
        milestone '16.10'
        restrict_gitlab_migration gitlab_schema: :gitlab_main
      
        MIGRATION = 'BackfillMergeRequestDiffsProjectId'
      
        def up
          ensure_batched_background_migration_is_finished(
            job_class_name: MIGRATION,
            table_name: :merge_request_diffs,
            column_name: :id,
            job_arguments: [],
            finalize: true
          )
        end
      
        def down
          # no-op
        end
      end
      
    • Add the NOT NULL constraint:

      # db/post_migrate/
      class AddMergeRequestDiffsProjectIdNotNullConstraint < Gitlab::Database::Migration[2.2]
        disable_ddl_transaction!
        milestone '16.7'
      
        def up
          add_not_null_constraint :merge_request_diffs, :project_id
        end
      
        def down
          remove_not_null_constraint :merge_request_diffs, :project_id
        end
      end
      
    • Optional. For very large tables, add an invalid NOT NULL constraint and schedule asynchronous validation:

      # db/post_migrate/
      class AddMergeRequestDiffsProjectIdNotNullConstraint < Gitlab::Database::Migration[2.2]
        disable_ddl_transaction!
        milestone '16.7'
      
        def up
          add_not_null_constraint :merge_request_diffs, :project_id, validate: false
        end
      
        def down
          remove_not_null_constraint :merge_request_diffs, :project_id
        end
      end
      
      # db/post_migrate/
      class PrepareMergeRequestDiffsProjectIdNotNullValidation < Gitlab::Database::Migration[2.2]
        milestone '16.10'
      
        CONSTRAINT_NAME = 'check_11c5f029ad'
      
        def up
          prepare_async_check_constraint_validation :merge_request_diffs, name: CONSTRAINT_NAME
        end
      
        def down
          unprepare_async_check_constraint_validation :merge_request_diffs, name: CONSTRAINT_NAME
        end
      end
      
    • Optional. For partitioned table, use:

      # db/post_migrate/
      
      PARTITIONED_TABLE_NAME = :p_ci_builds
      CONSTRAINT_NAME = 'check_9aa9432137'
      
      # Partitioned check constraint to be validated in https://gitlab.com/gitlab-org/gitlab/-/issues/XXXXX
      def up
        prepare_partitioned_async_check_constraint_validation PARTITIONED_TABLE_NAME, name: CONSTRAINT_NAME
      end
      
      def down
        unprepare_partitioned_async_check_constraint_validation PARTITIONED_TABLE_NAME, name: CONSTRAINT_NAME
      end
      
      note
      prepare_partitioned_async_check_constraint_validation only validates the existing NOT VALID check constraint asynchronously for all the partitions. It doesn’t create or validate the check constraint for the partitioned table.
  3. Optional. If the constraint was validated asynchronously, validate the NOT NULL constraint once validation is complete:

    # db/post_migrate/
    class ValidateMergeRequestDiffsProjectIdNullConstraint < Gitlab::Database::Migration[2.2]
      milestone '16.10'
    
      def up
        validate_not_null_constraint :merge_request_diffs, :project_id
      end
    
      def down
        # no-op
      end
    end
    

For these cases, consult the database team early in the update cycle. The NOT NULL constraint may not be required or other options could exist that do not affect really large or frequently accessed tables.

NOT NULL constraints for multiple columns

Sometimes we want to ensure a set of columns contains a specific number of NOT NULL values. A common example is a table that can belong to either a project or a group, and therefore project_id or group_id must be present. To enforce this, follow the steps for your use case above, but instead use the add_multi_column_not_null_constraint helper.

In this example, labels must belong to either a project or a group, but not both. We can add a check constraint to enforce this:

class AddLabelsNullConstraint < Gitlab::Database::Migration[2.2]
  disable_ddl_transaction!
  milestone '16.10'

  def up
    add_multi_column_not_null_constraint(:labels, :group_id, :project_id)
  end

  def down
    remove_multi_column_not_null_constraint(:labels, :group_id, :project_id)
  end
end

This will add the following constraint to labels:

CREATE TABLE labels (
    ...
    CONSTRAINT check_45e873b2a8 CHECK ((num_nonnulls(group_id, project_id) = 1))
);

num_nonnulls returns the number of supplied arguments that are non-null. Checking this value equals 1 in the constraint means that only one of group_id and project_id should contain a non-null value in a row, but not both.

Custom limits and operators

If we want to customize the number of non-nulls required, we can use a different limit and/or operator:

class AddLabelsNullConstraint < Gitlab::Database::Migration[2.2]
  disable_ddl_transaction!
  milestone '16.10'

  def up
    add_multi_column_not_null_constraint(:labels, :group_id, :project_id, limit: 0, operator: '>')
  end

  def down
    remove_multi_column_not_null_constraint(:labels, :group_id, :project_id)
  end
end

This is then reflected in the constraint, allowing both project_id and group_id to be present:

CREATE TABLE labels (
    ...
    CONSTRAINT check_45e873b2a8 CHECK ((num_nonnulls(group_id, project_id) > 0))
);