Database table partitioning

WARNING: If you have questions not answered below, check for and add them to this issue. Tag @gitlab-org/database-team/triage and we'll get back to you with an answer as soon as possible. If you get an answer in Slack, document it on the issue as well so we can update this document in the future.

Table partitioning is a powerful database feature that allows a table's data to be split into smaller physical tables that act as a single large table. If the application is designed to work with partitioning in mind, there can be multiple benefits, such as:

  • Query performance can be improved greatly, because the database can cheaply eliminate much of the data from the search space, while still providing full SQL capabilities.

  • Bulk deletes can be achieved with minimal impact on the database by dropping entire partitions. This is a natural fit for features that need to periodically delete data that falls outside the retention window.

  • Administrative tasks like VACUUM and index rebuilds can operate on individual partitions, rather than across a single massive table.

Unfortunately, not all models fit a partitioning scheme, and there are significant drawbacks if implemented incorrectly. Additionally, tables can only be partitioned at their creation, making it nontrivial to apply partitioning to a busy database. A suite of migration tools are available to enable backend developers to partition existing tables, but the migration process is rather heavy, taking multiple steps split across several releases. Due to the limitations of partitioning and the related migrations, you should understand how partitioning fits your use case before attempting to leverage this feature.

Determine when to use partitioning

While partitioning can be very useful when properly applied, it's imperative to identify if the data and workload of a table naturally fit a partitioning scheme. Understand a few details to decide if partitioning is a good fit for your particular problem:

  • Table partitioning. A table is partitioned on a partition key, which is a column or set of columns which determine how the data is split across the partitions. The partition key is used by the database when reading or writing data, to decide which partitions must be accessed. The partition key should be a column that would be included in a WHERE clause on almost all queries accessing that table.

  • How the data is split. What strategy does the database use to split the data across the partitions? The available choices are range, hash, and list.

Determine the appropriate partitioning strategy

The available partitioning strategy choices are range, hash, and list.

Range partitioning

The scheme best supported by the GitLab migration helpers is date-range partitioning, where each partition in the table contains data for a single month. In this case, the partitioning key must be a timestamp or date column. For this type of partitioning to work well, most queries must access data in a certain date range.

For a more concrete example, consider using the audit_events table. It was the first table to be partitioned in the application database (scheduled for deployment with the GitLab 13.5 release). This table tracks audit entries of security events that happen in the application. In almost all cases, users want to see audit activity that occurs in a certain time frame. As a result, date-range partitioning was a natural fit for how the data would be accessed.

To look at this in more detail, imagine a simplified audit_events schema:

CREATE TABLE audit_events (
  id SERIAL NOT NULL PRIMARY KEY,
  author_id INT NOT NULL,
  details jsonb NOT NULL,
  created_at timestamptz NOT NULL);

Now imagine typical queries in the UI would display the data in a certain date range, like a single week:

SELECT *
FROM audit_events
WHERE created_at >= '2020-01-01 00:00:00'
  AND created_at < '2020-01-08 00:00:00'
ORDER BY created_at DESC
LIMIT 100

If the table is partitioned on the created_at column the base table would look like:

CREATE TABLE audit_events (
  id SERIAL NOT NULL,
  author_id INT NOT NULL,
  details jsonb NOT NULL,
  created_at timestamptz NOT NULL,
  PRIMARY KEY (id, created_at))
PARTITION BY RANGE(created_at);

NOTE: The primary key of a partitioned table must include the partition key as part of the primary key definition.

And we might have a list of partitions for the table, such as:

audit_events_202001 FOR VALUES FROM ('2020-01-01') TO ('2020-02-01')
audit_events_202002 FOR VALUES FROM ('2020-02-01') TO ('2020-03-01')
audit_events_202003 FOR VALUES FROM ('2020-03-01') TO ('2020-04-01')

Each partition is a separate physical table, with the same structure as the base audit_events table, but contains only data for rows where the partition key falls in the specified range. For example, the partition audit_events_202001 contains rows where the created_at column is greater than or equal to 2020-01-01 and less than 2020-02-01.

Now, if we look at the previous example query again, the database can use the WHERE to recognize that all matching rows are in the audit_events_202001 partition. Rather than searching all of the data in all of the partitions, it can search only the single month's worth of data in the appropriate partition. In a large table, this can dramatically reduce the amount of data the database needs to access. However, imagine a query that does not filter based on the partitioning key, such as:

SELECT *
FROM audit_events
WHERE author_id = 123
ORDER BY created_at DESC
LIMIT 100

In this example, the database can't prune any partitions from the search, because matching data could exist in any of them. As a result, it has to query each partition individually, and aggregate the rows into a single result set. Because author_id would be indexed, the performance impact could likely be acceptable, but on more complex queries the overhead can be substantial. Partitioning should only be leveraged if the access patterns of the data support the partitioning strategy, otherwise performance suffers.

Hash Partitioning

Hash partitioning splits a logical table into a series of partitioned tables. Each partition corresponds to the ID range that matches a hash and remainder. For example, if partitioning BY HASH(id), rows with hash(id) % 64 == 1 would end up in the partition WITH (MODULUS 64, REMAINDER 1).

When hash partitioning, you must include a WHERE hashed_column = ? condition in every performance-sensitive query issued by the application. If this is not possible, hash partitioning may not be the correct fit for your use case.

Hash partitioning has one main advantage: it is the only type of partitioning that can enforce uniqueness on a single numeric id column. (While also possible with range partitioning, it's rarely the correct choice).

Hash partitioning has downsides:

  • The number of partitions must be known up-front.
  • It's difficult to move new data to an extra partition if current partitions become too large.
  • Range queries, such as WHERE id BETWEEN ? and ?, are unsupported.
  • Lookups by other keys, such as WHERE other_id = ?, are unsupported.

For this reason, it's often best to choose a large number of hash partitions to accommodate future table growth.

Partitioning a table (Range)

Unfortunately, tables can only be partitioned at their creation, making it nontrivial to apply to a busy database. A suite of migration tools have been developed to enable backend developers to partition existing tables. This migration process takes multiple steps which must be split across several releases.

Caveats

The partitioning migration helpers work by creating a partitioned duplicate of the original table and using a combination of a trigger and a background migration to copy data into the new table. Changes to the original table schema can be made in parallel with the partitioning migration, but they must take care to not break the underlying mechanism that makes the migration work. For example, if a column is added to the table that is being partitioned, both the partitioned table and the trigger definition must be updated to match.

Step 1: Creating the partitioned copy (Release N)

The first step is to add a migration to create the partitioned copy of the original table. This migration creates the appropriate partitions based on the data in the original table, and install a trigger that syncs writes from the original table into the partitioned copy.

An example migration of partitioning the audit_events table by its created_at column would look like:

class PartitionAuditEvents < Gitlab::Database::Migration[2.1]
  include Gitlab::Database::PartitioningMigrationHelpers

  def up
    partition_table_by_date :audit_events, :created_at
  end

  def down
    drop_partitioned_table_for :audit_events
  end
end

After this has executed, any inserts, updates, or deletes in the original table are also duplicated in the new table. For updates and deletes, the operation only has an effect if the corresponding row exists in the partitioned table.

Step 2: Backfill the partitioned copy (Release N)

The second step is to add a post-deployment migration that schedules the background jobs that backfill existing data from the original table into the partitioned copy.

Continuing the above example, the migration would look like:

class BackfillPartitionAuditEvents < Gitlab::Database::Migration[2.1]
  include Gitlab::Database::PartitioningMigrationHelpers

  def up
    enqueue_partitioning_data_migration :audit_events
  end

  def down
    cleanup_partitioning_data_migration :audit_events
  end
end

This step uses the same mechanism as any background migration, so you may want to read the Background Migration guide for details on that process. Background jobs are scheduled every 2 minutes and copy 50_000 records at a time, which can be used to estimate the timing of the background migration portion of the partitioning migration.

Step 3: Post-backfill cleanup (Release N+1)

The third step must occur at least one release after the release that includes the background migration. This gives time for the background migration to execute properly in self-managed installations. In this step, add another post-deployment migration that cleans up after the background migration. This includes forcing any remaining jobs to execute, and copying data that may have been missed, due to dropped or failed jobs.

Once again, continuing the example, this migration would look like:

class CleanupPartitionedAuditEventsBackfill < Gitlab::Database::Migration[2.1]
  include Gitlab::Database::PartitioningMigrationHelpers

  def up
    finalize_backfilling_partitioned_table :audit_events
  end

  def down
    # no op
  end
end

After this migration has completed, the original table and partitioned table should contain identical data. The trigger installed on the original table guarantees that the data remains in sync going forward.

Step 4: Swap the partitioned and non-partitioned tables (Release N+1)

The final step of the migration makes the partitioned table ready for use by the application. This section will be updated when the migration helper is ready, for now development can be followed in the Tracking Issue.

Partitioning a table (Hash)

Hash partitioning divides data into partitions based on a hash of their ID. It works well only if most queries against the table include a clause like WHERE id = ?, so that PostgreSQL can decide which partition to look in based on the ID or ids being requested.

Another key downside is that hash partitioning does not allow adding additional partitions after table creation. The correct number of partitions must be chosen up-front.

Hash partitioning is the only type of partitioning (aside from some complex uses of list partitioning) that can guarantee uniqueness of an ID across multiple partitions at the database level.

Partitioning a table (List)

Introduced in GitLab 15.4.

Add the partitioning key column to the table you are partitioning. Include the partitioning key in the following constraints:

  • The primary key.
  • All foreign keys referencing the table to be partitioned.
  • All unique constraints.

Step 1 - Add partition key

Add the partitioning key column. For example, in a rails migration:

class AddPartitionNumberForPartitioning < Gitlab::Database::Migration[2.1]
  enable_lock_retries!

  TABLE_NAME = :table_name
  COLUMN_NAME = :partition_id
  DEFAULT_VALUE = 100

  def change
    add_column(TABLE_NAME, COLUMN_NAME, :bigint, default: 100)
  end
end

Step 2 - Create required indexes

Add indexes including the partitioning key column. For example, in a rails migration:

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

  TABLE_NAME = :table_name
  INDEX_NAME = :index_name

  def up
    add_concurrent_index(TABLE_NAME, [:id, :partition_id], unique: true, name: INDEX_NAME)
  end

  def down
    remove_concurrent_index_by_name(TABLE_NAME, INDEX_NAME)
  end
end

Step 3 - Enforce unique constraint

Change all unique indexes to include the partitioning key column, including the primary key index. You can start by adding an unique index on [primary_key_column, :partition_id], which will be required for the next two steps. For example, in a rails migration:

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

  TABLE_NAME = :table_name
  OLD_UNIQUE_INDEX_NAME = :index_name_unique
  NEW_UNIQUE_INDEX_NAME = :new_index_name

  def up
    add_concurrent_index(TABLE_NAME, [:id, :partition_id], unique: true, name: NEW_UNIQUE_INDEX_NAME)

    remove_concurrent_index_by_name(TABLE_NAME, OLD_UNIQUE_INDEX_NAME)
  end

  def down
    add_concurrent_index(TABLE_NAME, :id, unique: true, name: OLD_UNIQUE_INDEX_NAME)

    remove_concurrent_index_by_name(TABLE_NAME, NEW_UNIQUE_INDEX_NAME)
  end
end

Step 4 - Enforce foreign key constraint

Enforce foreign keys including the partitioning key column. For example, in a rails migration:

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

  SOURCE_TABLE_NAME = :source_table_name
  TARGET_TABLE_NAME = :target_table_name
  COLUMN = :foreign_key_id
  TARGET_COLUMN = :id
  FK_NAME = :fk_365d1db505_p
  PARTITION_COLUMN = :partition_id

  def up
    add_concurrent_foreign_key(
      SOURCE_TABLE_NAME,
      TARGET_TABLE_NAME,
      column: [PARTITION_COLUMN, COLUMN],
      target_column: [PARTITION_COLUMN, TARGET_COLUMN],
      validate: false,
      on_update: :cascade,
      name: FK_NAME
    )

    # This should be done in a separate post migration when dealing with a high traffic table
    validate_foreign_key(TABLE_NAME, [PARTITION_COLUMN, COLUMN], name: FK_NAME)
  end

  def down
    with_lock_retries do
      remove_foreign_key_if_exists(SOURCE_TABLE_NAME, name: FK_NAME)
    end
  end
end

The on_update: :cascade option is mandatory if we want the partitioning column to be updated. This will cascade the update to all dependent rows. Without specifying it, updating the partition column on the target table we would result in a Key is still referenced from table ... error and updating the partition column on the source table would raise a Key is not present in table ... error.

This migration can be automatically generated using:

./scripts/partitioning/generate-fk --source source_table_name --target target_table_name

Step 5 - Swap primary key

Swap the primary key including the partitioning key column. This can be done only after including the partition key for all references foreign keys. For example, in a rails migration:

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

  TABLE_NAME = :table_name
  PRIMARY_KEY = :primary_key
  OLD_INDEX_NAME = :old_index_name
  NEW_INDEX_NAME = :new_index_name

  def up
    swap_primary_key(TABLE_NAME, PRIMARY_KEY, NEW_INDEX_NAME)
  end

  def down
    add_concurrent_index(TABLE_NAME, :id, unique: true, name: OLD_INDEX_NAME)
    add_concurrent_index(TABLE_NAME, [:id, :partition_id], unique: true, name: NEW_INDEX_NAME)

    unswap_primary_key(TABLE_NAME, PRIMARY_KEY, OLD_INDEX_NAME)
  end
end

NOTE: Do not forget to set the primary key explicitly in your model as ActiveRecord does not support composite primary keys.

class Model < ApplicationRecord
  self.primary_key = :id
end

Step 6 - Create parent table and attach existing table as the initial partition

You can now create the parent table attaching the existing table as the initial partition by using the following helpers provided by the database team.

For example, using list partitioning in Rails post migrations:

class PrepareTableConstraintsForListPartitioning < Gitlab::Database::Migration[2.1]
  include Gitlab::Database::PartitioningMigrationHelpers::TableManagementHelpers

  disable_ddl_transaction!

  TABLE_NAME = :table_name
  PARENT_TABLE_NAME = :p_table_name
  FIRST_PARTITION = 100
  PARTITION_COLUMN = :partition_id

  def up
    prepare_constraint_for_list_partitioning(
      table_name: TABLE_NAME,
      partitioning_column: PARTITION_COLUMN,
      parent_table_name: PARENT_TABLE_NAME,
      initial_partitioning_value: FIRST_PARTITION
    )
  end

  def down
    revert_preparing_constraint_for_list_partitioning(
      table_name: TABLE_NAME,
      partitioning_column: PARTITION_COLUMN,
      parent_table_name: PARENT_TABLE_NAME,
      initial_partitioning_value: FIRST_PARTITION
    )
  end
end
class ConvertTableToListPartitioning < Gitlab::Database::Migration[2.1]
  include Gitlab::Database::PartitioningMigrationHelpers::TableManagementHelpers

  disable_ddl_transaction!

  TABLE_NAME = :table_name
  TABLE_FK = :table_references_by_fk
  PARENT_TABLE_NAME = :p_table_name
  FIRST_PARTITION = 100
  PARTITION_COLUMN = :partition_id

  def up
    convert_table_to_first_list_partition(
      table_name: TABLE_NAME,
      partitioning_column: PARTITION_COLUMN,
      parent_table_name: PARENT_TABLE_NAME,
      initial_partitioning_value: FIRST_PARTITION,
      lock_tables: [TABLE_FK, TABLE_NAME]
    )
  end

  def down
    revert_converting_table_to_first_list_partition(
      table_name: TABLE_NAME,
      partitioning_column: PARTITION_COLUMN,
      parent_table_name: PARENT_TABLE_NAME,
      initial_partitioning_value: FIRST_PARTITION
    )
  end
end

NOTE: Do not forget to set the sequence name explicitly in your model because it will be owned by the routing table and ActiveRecord can't determine it. This can be cleaned up after the table_name is changed to the routing table.

class Model < ApplicationRecord
  self.sequence_name = 'model_id_seq'
end

If the partitioning constraint migration takes more than 10 minutes to finish, it can be made to run asynchronously to avoid running the post-migration during busy hours.

Prepend the following migration AsyncPrepareTableConstraintsForListPartitioning and use async: true option. This change marks the partitioning constraint as NOT VALID and enqueues a scheduled job to validate the existing data in the table during the weekend.

Then the second post-migration PrepareTableConstraintsForListPartitioning only marks the partitioning constraint as validated, because the existing data is already tested during the previous weekend.

For example:

class AsyncPrepareTableConstraintsForListPartitioning < Gitlab::Database::Migration[2.1]
  include Gitlab::Database::PartitioningMigrationHelpers::TableManagementHelpers

  disable_ddl_transaction!

  TABLE_NAME = :table_name
  PARENT_TABLE_NAME = :p_table_name
  FIRST_PARTITION = 100
  PARTITION_COLUMN = :partition_id

  def up
    prepare_constraint_for_list_partitioning(
      table_name: TABLE_NAME,
      partitioning_column: PARTITION_COLUMN,
      parent_table_name: PARENT_TABLE_NAME,
      initial_partitioning_value: FIRST_PARTITION,
      async: true
    )
  end

  def down
    revert_preparing_constraint_for_list_partitioning(
      table_name: TABLE_NAME,
      partitioning_column: PARTITION_COLUMN,
      parent_table_name: PARENT_TABLE_NAME,
      initial_partitioning_value: FIRST_PARTITION
    )
  end
end