Adding Database Indexes

Indexes can be used to speed up database queries, but when should you add a new index? Traditionally the answer to this question has been to add an index for every column used for filtering or joining data. For example, consider the following query:

SELECT *
FROM projects
WHERE user_id = 2;

Here we are filtering by the user_id column and as such a developer may decide to index this column.

While in certain cases indexing columns using the above approach may make sense it can actually have a negative impact. Whenever you write data to a table any existing indexes need to be updated. The more indexes there are the slower this can potentially become. Indexes can also take up quite some disk space depending on the amount of data indexed and the index type. For example, PostgreSQL offers "GIN" indexes which can be used to index certain data types that can not be indexed by regular B-tree indexes. These indexes however generally take up more data and are slower to update compared to B-tree indexes.

Because of all this one should not blindly add a new index for every column used to filter data by. Instead one should ask themselves the following questions:

  1. Can I write my query in such a way that it re-uses as many existing indexes as possible?
  2. Is the data going to be large enough that using an index will actually be faster than just iterating over the rows in the table?
  3. Is the overhead of maintaining the index worth the reduction in query timings?

We'll explore every question in detail below.

Re-using Queries

The first step is to make sure your query re-uses as many existing indexes as possible. For example, consider the following query:

SELECT *
FROM todos
WHERE user_id = 123
AND state = 'open';

Now imagine we already have an index on the user_id column but not on the state column. One may think this query will perform badly due to state being unindexed. In reality the query may perform just fine given the index on user_id can filter out enough rows.

The best way to determine if indexes are re-used is to run your query using EXPLAIN ANALYZE. Depending on any extra tables that may be joined and other columns being used for filtering you may find an extra index is not going to make much (if any) difference. On the other hand you may determine that the index may make a difference.

In short:

  1. Try to write your query in such a way that it re-uses as many existing indexes as possible.
  2. Run the query using EXPLAIN ANALYZE and study the output to find the most ideal query.

Data Size

A database may decide not to use an index despite it existing in case a regular sequence scan (= simply iterating over all existing rows) is faster. This is especially the case for small tables.

If a table is expected to grow in size and you expect your query has to filter out a lot of rows you may want to consider adding an index. If the table size is very small (for example, fewer than 1,000 records) or any existing indexes filter out enough rows you may not want to add a new index.

Maintenance Overhead

Indexes have to be updated on every table write. In case of PostgreSQL all existing indexes will be updated whenever data is written to a table. As a result of this having many indexes on the same table will slow down writes.

Because of this one should ask themselves: is the reduction in query performance worth the overhead of maintaining an extra index?

If adding an index reduces SELECT timings by 5 milliseconds but increases INSERT/UPDATE/DELETE timings by 10 milliseconds then the index may not be worth it. On the other hand, if SELECT timings are reduced but INSERT/UPDATE/DELETE timings are not affected you may want to add the index after all.

Finding Unused Indexes

To see which indexes are unused you can run the following query:

SELECT relname as table_name, indexrelname as index_name, idx_scan, idx_tup_read, idx_tup_fetch, pg_size_pretty(pg_relation_size(indexrelname::regclass))
FROM pg_stat_all_indexes
WHERE schemaname = 'public'
AND idx_scan = 0
AND idx_tup_read = 0
AND idx_tup_fetch = 0
ORDER BY pg_relation_size(indexrelname::regclass) desc;

This query outputs a list containing all indexes that are never used and sorts them by indexes sizes in descending order. This query can be useful to determine if any previously indexes are useful after all. More information on the meaning of the various columns can be found at https://www.postgresql.org/docs/current/monitoring-stats.html.

Because the output of this query relies on the actual usage of your database it may be affected by factors such as (but not limited to):

  • Certain queries never being executed, thus not being able to use certain indexes.
  • Certain tables having little data, resulting in PostgreSQL using sequence scans instead of index scans.

In other words, this data is only reliable for a frequently used database with plenty of data and with as many GitLab features enabled (and being used) as possible.

Requirements for naming indexes

Indexes with complex definitions need to be explicitly named rather than relying on the implicit naming behavior of migration methods. In short, that means you must provide an explicit name argument for an index created with one or more of the following options:

  • where
  • using
  • order
  • length
  • type
  • opclass

Considerations for index names

Index names don't have any significance in the database, so they should attempt to communicate intent to others. The most important rule to remember is that generic names are more likely to conflict or be duplicated, and should not be used. Some other points to consider:

  • For general indexes, use a template, like: index_{table}_{column}_{options}.
  • For indexes added to solve a very specific problem, it may make sense for the name to reflect their use.
  • Identifiers in PostgreSQL have a maximum length of 63 bytes.
  • Check db/structure.sql for conflicts and ideas.

Why explicit names are required

As Rails is database agnostic, it generates an index name only from the required options of all indexes: table name and column name(s). For example, imagine the following two indexes are created in a migration:

def up
  add_index :my_table, :my_column

  add_index :my_table, :my_column, where: 'my_column IS NOT NULL'
end

Creation of the second index would fail, because Rails would generate the same name for both indexes.

This is further complicated by the behavior of the index_exists? method. It considers only the table name, column name(s) and uniqueness specification of the index when making a comparison. Consider:

def up
  unless index_exists?(:my_table, :my_column, where: 'my_column IS NOT NULL')
    add_index :my_table, :my_column, where: 'my_column IS NOT NULL'
  end
end

The call to index_exists? will return true if any index exists on :my_table and :my_column, and index creation will be bypassed.

The add_concurrent_index helper is a requirement for creating indexes on populated tables. Since it cannot be used inside a transactional migration, it has a built-in check that detects if the index already exists. In the event a match is found, index creation is skipped. Without an explicit name argument, Rails can return a false positive for index_exists?, causing a required index to not be created properly. By always requiring a name for certain types of indexes, the chance of error is greatly reduced.

Temporary indexes

There may be times when an index is only needed temporarily.

For example, in a migration, a column of a table might be conditionally updated. To query which columns need to be updated within the query performance guidelines, an index is needed that would otherwise not be used.

In these cases, a temporary index should be considered. To specify a temporary index:

  1. Prefix the index name with tmp_ and follow the naming conventions and requirements for naming indexes for the rest of the name.
  2. Create a follow-up issue to remove the index in the next (or future) milestone.
  3. Add a comment in the migration mentioning the removal issue.

A temporary migration would look like:

INDEX_NAME = 'tmp_index_projects_on_owner_where_emails_disabled'

def up
  # Temporary index to be removed in 13.9 https://gitlab.com/gitlab-org/gitlab/-/issues/1234
  add_concurrent_index :projects, :creator_id, where: 'emails_disabled = false', name: INDEX_NAME
end

def down
  remove_concurrent_index_by_name :projects, INDEX_NAME
end

Create indexes asynchronously

For very large tables, index creation can be a challenge to manage. While add_concurrent_index creates indexes in a way that does not block normal traffic, it can still be problematic when index creation runs for many hours. Necessary database operations like autovacuum cannot run, and on GitLab.com, the deployment process is blocked waiting for index creation to finish.

To limit impact on GitLab.com, a process exists to create indexes asynchronously during weekend hours. Due to generally lower levels of traffic and lack of regular deployments, this process allows the creation of indexes to proceed with a lower level of risk. The below sections describe the steps required to use these features:

  1. Schedule the index to be created.
  2. Verify the MR was deployed and the index exists in production.
  3. Add a migration to create the index synchronously.

Schedule the index to be created

Create an MR with a post-deployment migration which prepares the index for asynchronous creation. An example of creating an index using the asynchronous index helpers can be seen in the block below. This migration enters the index name and definition into the postgres_async_indexes table. The process that runs on weekends pulls indexes from this table and attempt to create them.

# in db/post_migrate/

INDEX_NAME = 'index_ci_builds_on_some_column'

def up
  prepare_async_index :ci_builds, :some_column, name: INDEX_NAME
end

def down
  unprepare_async_index :ci_builds, :some_column, name: INDEX_NAME
end

Verify the MR was deployed and the index exists in production

You can verify if the MR was deployed to GitLab.com by executing /chatops run auto_deploy status <merge_sha>. To verify existence of the index, you can:

  • Use a meta-command in #database-lab, such as: \d <index_name>
    • Ensure that the index is not invalid
  • Ask someone in #database to check if the index exists
  • With proper access, you can also verify directly on production or in a production clone

Add a migration to create the index synchronously

After the index is verified to exist on the production database, create a second merge request that adds the index synchronously. The synchronous migration results in a no-op on GitLab.com, but you should still add the migration as expected for other installations. The below block demonstrates how to create the second migration for the previous asynchronous example.

WARNING: The responsibility lies on the individual writing the migrations to verify the index exists in production before merging a second migration that adds the index using add_concurrent_index. If the second migration is deployed and the index has not yet been created, the index is created synchronously when the second migration executes.

# in db/post_migrate/

INDEX_NAME = 'index_ci_builds_on_some_column'

disable_ddl_transaction!

def up
  add_concurrent_index :ci_builds, :some_column, name: INDEX_NAME
end

def down
  remove_concurrent_index_by_name :ci_builds, INDEX_NAME
end

Test database index changes locally

You must test the database index changes locally before creating a merge request.

Verify indexes created asynchronously

Use the asynchronous index helpers on your local environment to test changes for creating an index:

  1. Enable the feature flags by running Feature.enable(:database_async_index_creation) and Feature.enable(:database_reindexing) in the Rails console.
  2. Run bundle exec rails db:migrate so that it creates an entry in the postgres_async_indexes table.
  3. Run bundle exec rails gitlab:db:reindex so that the index is created asynchronously.
  4. To verify the index, open the PostgreSQL console using the GDK command gdk psql and run the command \d <index_name> to check that your newly created index exists.