SQL Tuning with EXPLAIN

On this page Carat arrow pointing down
Warning:
CockroachDB v21.1 is no longer supported as of November 18, 2022. For more details, refer to the Release Support Policy.

This tutorial guides you through the common reasons for slow SQL statements and describes how to use EXPLAIN to troubleshoot the issues.

The following examples use MovR, a fictional vehicle-sharing application, to demonstrate CockroachDB SQL statements. Run cockroach demo movr to open an interactive SQL shell to a temporary, in-memory cluster with the movr database preloaded and set as the current database.

Issue: Full table scans

The most common reason for slow queries is sub-optimal SELECT statements that include full table scans and incorrect use of indexes.

You'll get generally poor performance when retrieving a single row based on a column that is not in the primary key or any secondary index:

icon/buttons/copy
> SELECT * FROM users WHERE name = 'Cheyenne Smith';
                   id                |   city   |      name      |      address      | credit_card
-------------------------------------+----------+----------------+-------------------+--------------
00e6afcc-e1c5-4258-8000-00000000002c | new york | Cheyenne Smith | 8550 Kelsey Flats | 4374468739
(1 row)

Time: 14ms total (execution 14ms / network 0ms)

To understand why this query performs poorly, use EXPLAIN:

icon/buttons/copy
> EXPLAIN SELECT * FROM users WHERE name = 'Cheyenne Smith';
                                          info
-----------------------------------------------------------------------------------------
  distribution: full
  vectorized: true

  • filter
  │ estimated row count: 1
  │ filter: name = 'Cheyenne Smith'
  │
  └── • scan
        estimated row count: 12,500 (100% of the table; stats collected 56 minutes ago)
        table: users@primary
        spans: FULL SCAN
(11 rows)

Time: 2ms total (execution 1ms / network 2ms)

table: users@primary indicates the index used (primary) to scan the table (users). spans: FULL SCAN shows you that, without a secondary index on the name column, CockroachDB scans every row of the users table, ordered by the primary key (city/id), until it finds the row with the correct name value.

Solution: Filter by a secondary index

To speed up this query, add a secondary index on name:

icon/buttons/copy
> CREATE INDEX on users (name);

The query will now return much faster:

icon/buttons/copy
> SELECT * FROM users WHERE name = 'Cheyenne Smith';
                   id                  |   city   |      name      |      address      | credit_card
---------------------------------------+----------+----------------+-------------------+--------------
  00e6afcc-e1c5-4258-8000-00000000002c | new york | Cheyenne Smith | 8550 Kelsey Flats | 4374468739
(1 row)

Time: 7ms total (execution 7ms / network 0ms)

To understand why the performance improved, use EXPLAIN to see the new query plan:

icon/buttons/copy
> EXPLAIN SELECT * FROM users WHERE name = 'Cheyenne Smith';
                                         info
--------------------------------------------------------------------------------------
  distribution: local
  vectorized: true

  • index join
  │ estimated row count: 1
  │ table: users@primary
  │
  └── • scan
        estimated row count: 1 (<0.01% of the table; stats collected 58 minutes ago)
        table: users@users_name_idx
        spans: [/'Cheyenne Smith' - /'Cheyenne Smith']
(11 rows)

Time: 1ms total (execution 1ms / network 0ms)

This shows you that CockroachDB starts with the secondary index (users@users_name_idx). Because it is sorted by name, the query can jump directly to the relevant value (/'Cheyenne Smith' - /'Cheyenne Smith'). However, the query needs to return values not in the secondary index, so CockroachDB grabs the primary key (city/id) stored with the name value (the primary key is always stored with entries in a secondary index), jumps to that value in the primary index, and then returns the full row.

Because the users table is under 512 MiB, the primary index and all secondary indexes are contained in a single range with a single leaseholder. If the table were bigger, however, the primary index and secondary index could reside in separate ranges, each with its own leaseholder. In this case, if the leaseholders were on different nodes, the query would require more network hops, further increasing latency.

Solution: Filter by a secondary index storing additional columns

When you have a query that filters by a specific column but retrieves a subset of the table's total columns, you can improve performance by storing those additional columns in the secondary index to prevent the query from needing to scan the primary index as well.

For example, let's say you frequently retrieve a user's name and credit card number:

icon/buttons/copy
> SELECT name, credit_card FROM users WHERE name = 'Cheyenne Smith';
       name      | credit_card
-----------------+--------------
  Cheyenne Smith | 4374468739
(1 row)

Time: 6ms total (execution 6ms / network 0ms)

With the current secondary index on name, CockroachDB still needs to scan the primary index to get the credit card number:

icon/buttons/copy
> EXPLAIN SELECT name, credit_card FROM users WHERE name = 'Cheyenne Smith';
                                        info
-------------------------------------------------------------------------------------
  distribution: local
  vectorized: true

  • index join
  │ estimated row count: 1
  │ table: users@primary
  │
  └── • scan
        estimated row count: 1 (<0.01% of the table; stats collected 2 minutes ago)
        table: users@users_name_idx
        spans: [/'Cheyenne Smith' - /'Cheyenne Smith']
(11 rows)

Time: 1ms total (execution 1ms / network 0ms)

Let's drop and recreate the index on name, this time storing the credit_card value in the index:

icon/buttons/copy
> DROP INDEX users_name_idx;
icon/buttons/copy
> CREATE INDEX ON users (name) STORING (credit_card);

Now that credit_card values are stored in the index on name, CockroachDB only needs to scan that index:

icon/buttons/copy
> EXPLAIN SELECT name, credit_card FROM users WHERE name = 'Cheyenne Smith';
                                      info
---------------------------------------------------------------------------------
  distribution: local
  vectorized: true

  • scan
    estimated row count: 1 (<0.01% of the table; stats collected 2 minutes ago)
    table: users@users_name_idx
    spans: [/'Cheyenne Smith' - /'Cheyenne Smith']
(7 rows)

Time: 1ms total (execution 1ms / network 0ms)

This results in even faster performance:

icon/buttons/copy
> SELECT name, credit_card FROM users WHERE name = 'Cheyenne Smith';
name      | credit_card
-----------------+--------------
Cheyenne Smith | 4374468739
(1 row)

Time: 1ms total (execution 1ms / network 0ms)

To reset the database for following examples, let's drop the index on name:

icon/buttons/copy
> DROP INDEX users_name_idx;

Issue: Joining data from different tables

Secondary indexes are crucial when joining data from different tables as well.

For example, let's say you want to count the number of users who started rides on a given day. To do this, you need to use a join to get the relevant rides from the rides table and then map the rider_id for each of those rides to the corresponding id in the users table, counting each mapping only once:

icon/buttons/copy
SELECT count(DISTINCT users.id) FROM users INNER JOIN rides ON rides.rider_id = users.id WHERE start_time BETWEEN '2018-12-16 00:00:00' AND '2018-12-17 00:00:00';
  count
---------
   17
(1 row)

Time: 4ms total (execution 3ms / network 0ms)

To understand what's happening, use EXPLAIN to see the query plan:

icon/buttons/copy
> EXPLAIN SELECT count(DISTINCT users.id) FROM users INNER JOIN rides ON rides.rider_id = users.id WHERE start_time BETWEEN '2018-12-16 00:00:00' AND '2018-12-17 00:00:00';
                                                  info
---------------------------------------------------------------------------------------------------------
  distribution: full
  vectorized: true

  • group (scalar)
  │ estimated row count: 1
  │
  └── • distinct
      │ estimated row count: 14
      │ distinct on: id
      │
      └── • hash join
          │ estimated row count: 16
          │ equality: (id) = (rider_id)
          │
          ├── • scan
          │     estimated row count: 50 (100% of the table; stats collected 2 minutes ago)
          │     table: users@primary
          │     spans: FULL SCAN
          │
          └── • filter
              │ estimated row count: 16
              │ filter: (start_time >= '2018-12-16 00:00:00') AND (start_time <= '2018-12-17 00:00:00')
              │
              └── • scan
                    estimated row count: 500 (100% of the table; stats collected 2 minutes ago)
                    table: rides@primary
                    spans: FULL SCAN
(27 rows)

Time: 1ms total (execution 1ms / network 0ms)

CockroachDB does a full table scan first on rides to get all rows with a start_time in the specified range and then does another full table scan on users to find matching rows and calculate the count.

Given the WHERE condition of the join, the full table scan of rides is particularly wasteful.

Solution: Create a secondary index on the WHERE condition storing the join key

To speed up the query, you can create a secondary index on the WHERE condition (rides.start_time) storing the join key (rides.rider_id):

icon/buttons/copy
> CREATE INDEX ON rides (start_time) STORING (rider_id);

Adding the secondary index reduced the query time:

icon/buttons/copy
SELECT count(DISTINCT users.id) FROM users INNER JOIN rides ON rides.rider_id = users.id WHERE start_time BETWEEN '2018-12-16 00:00:00' AND '2018-12-17 00:00:00';
  count
---------
   17
(1 row)

Time: 2ms total (execution 2ms / network 0ms)

To understand why performance improved, again use EXPLAIN to see the new query plan:

icon/buttons/copy
> EXPLAIN SELECT count(DISTINCT users.id) FROM users INNER JOIN rides ON rides.rider_id = users.id WHERE start_time BETWEEN '2020-09-16 00:00:00' AND '2020-09-17 00:00:00';
                                            info
--------------------------------------------------------------------------------------------
  distribution: full
  vectorized: true

  • group (scalar)
  │ estimated row count: 1
  │
  └── • distinct
      │ estimated row count: 14
      │ distinct on: id
      │
      └── • hash join
          │ estimated row count: 16
          │ equality: (id) = (rider_id)
          │
          ├── • scan
          │     estimated row count: 50 (100% of the table; stats collected 4 minutes ago)
          │     table: users@primary
          │     spans: FULL SCAN
          │
          └── • scan
                estimated row count: 16 (3.2% of the table; stats collected 4 minutes ago)
                table: rides@rides_start_time_idx
                spans: [/'2018-12-16 00:00:00' - /'2018-12-17 00:00:00']
(23 rows)

Time: 1ms total (execution 1ms / network 0ms)

Notice that CockroachDB now starts by using rides@rides_start_time_idx secondary index to retrieve the relevant rides without needing to scan the full rides table.

Issue: Inefficient joins

Hash joins are more expensive and require more memory than lookup joins. Hence the cost-based optimizer uses a lookup join whenever possible.

For the following query, the cost-based optimizer can’t perform a lookup join because the query doesn’t have a prefix of the rides table’s primary key available and thus has to read the entire table and search for a match, resulting in a slow query:

icon/buttons/copy
> EXPLAIN SELECT * FROM vehicles JOIN rides on rides.vehicle_id = vehicles.id limit 1;
------------------------------------------------------------------------------------------
  distribution: full
  vectorized: true

  • limit
  │ estimated row count: 1
  │ count: 1
  │
  └── • hash join
      │ estimated row count: 500
      │ equality: (vehicle_id) = (id)
      │
      ├── • scan
      │     estimated row count: 500 (100% of the table; stats collected 23 seconds ago)
      │     table: rides@primary
      │     spans: FULL SCAN
      │
      └── • scan
            estimated row count: 15 (100% of the table; stats collected 4 minutes ago)
            table: vehicles@primary
            spans: FULL SCAN
(20 rows)

Time: 1ms total (execution 1ms / network 0ms)

Solution: Provide primary key to allow lookup join

To speed up the query, you can provide the primary key to allow the cost-based optimizer to perform a lookup join instead of a hash join:

icon/buttons/copy
> EXPLAIN SELECT * FROM vehicles JOIN rides ON rides.vehicle_id = vehicles.id and rides.city = vehicles.city limit 1;
                                          info
----------------------------------------------------------------------------------------
  distribution: full
  vectorized: true

  • limit
  │ estimated row count: 1
  │ count: 1
  │
  └── • lookup join
      │ estimated row count: 56
      │ table: vehicles@primary
      │ equality: (city, vehicle_id) = (city,id)
      │ equality cols are key
      │
      └── • scan
            estimated row count: 500 (100% of the table; stats collected 1 minute ago)
            table: rides@primary
            spans: FULL SCAN
(17 rows)

Time: 1ms total (execution 1ms / network 0ms)

See also


Yes No
On this page

Yes No