SHOW ANALYZE

MariaDB starting with 10.9

SHOW ANALYZE was added in MariaDB 10.9.

Syntax

SHOW ANALYZE [FORMAT=JSON] FOR <connection_id>;

Description

SHOW ANALYZE allows one to retrieve ANALYZE-like output from a currently running statement. The command

SHOW ANALYZE [FORMAT=JSON] FOR <connection_id>;

connects to the query running in connection connection_id, gets information about the query plan it is executing, also gets information about the runtime statistics of the execution so far and returns it in a format similar to ANALYZE [FORMAT=JSON] output.

This is similar to the SHOW EXPLAIN command, the difference being that SHOW ANALYZE also produces runtime statistics information.

Intended Usage

Imagine you're trying to troubleshoot a query that never finishes. Since it doesn't finish, it is not possible to get ANALYZE output for it. With SHOW ANALYZE, you can get the runtime statistics without waiting for the query to finish.

Examples

Example 1: Row Counts

Consider the tables orders and customer and a join query finding the total amount of orders from customers with Gold status:

explain format=json
select sum(orders.amount)
from
  customer join orders on customer.cust_id=orders.cust_id
where
  customer.status='GOLD';

The EXPLAIN for this query looks like this:

+------+-------------+----------+------+---------------+---------+---------+------------------+--------+-------------+
| id   | select_type | table    | type | possible_keys | key     | key_len | ref              | rows   | Extra       |
+------+-------------+----------+------+---------------+---------+---------+------------------+--------+-------------+
|    1 | SIMPLE      | customer | ALL  | PRIMARY       | NULL    | NULL    | NULL             | 199786 | Using where |
|    1 | SIMPLE      | orders   | ref  | cust_id       | cust_id | 5       | customer.cust_id | 1      |             |
+------+-------------+----------+------+---------------+---------+---------+------------------+--------+-------------+

We run the SELECT, and it has been running for 30 seconds. Let's try SHOW ANALYZE:

show analyze format=json for 3;
| {
  "r_query_time_in_progress_ms": 32138,

^ this shows how long the query has been running.

  "query_block": {
    "select_id": 1,
    "r_loops": 1,
    "nested_loop": [
      {
        "table": {
          "table_name": "customer",
          "access_type": "ALL",
          "possible_keys": ["PRIMARY"],
          "r_loops": 1,
          "rows": 199786,
          "r_rows": 110544,

^ rows shows the number of rows expected. r_rows in this example shows how many rows were processed so far (110K out of expected 200K). r_loops above shows we're doing the first table scan (which is obvious for this query plan).

          "filtered": 100,
          "r_filtered": 9.538283398,
          "attached_condition": "customer.`status` = 'GOLD'"
        }
      },
      {
        "table": {
          "table_name": "orders",
          "access_type": "ref",
          "possible_keys": ["cust_id"],
          "key": "cust_id",
          "key_length": "5",
          "used_key_parts": ["cust_id"],
          "ref": ["test.customer.cust_id"],
          "r_loops": 10544,
          "rows": 1,
          "r_rows": 99.99222307,

^ here, rows: 1 shows the optimizer was expecting 1 order per customer. But r_rows: 99.9 shows that execution has found on average 100 orders per customer. This may be the reason the query is slower than expected!

The final chunk of the output doesn't have anything interesting but here it is:

          "filtered": 100,
          "r_filtered": 100
        }
      }
    ]
  }
}

Example 2: Timing Information

Regular SELECT queries collect row count information, so SHOW ANALYZE can display it. However, detailed timing information is not collected, as collecting it may have CPU overhead. But if the target query is collecting timing information, SHOW ANALYZE will display it. How does one get the target query to collect timing information? Currently there is one way: if the target is running ANALYZE, it IS collecting timing information. Re-running the previous example:

Connection 1> ANALYZE SELECT ... ; 
Connection 2> SHOW ANALYZE FORMAT=JSON FOR <connection_id>;
ANALYZE
{
  "r_query_time_in_progress_ms": 30727,
  "query_block": {
    "select_id": 1,
    "r_loops": 1,
    "nested_loop": [
      {
        "table": {
          "table_name": "customer",
          "access_type": "ALL",
          "possible_keys": ["PRIMARY"],
          "r_loops": 1,
          "rows": 199786,
          "r_rows": 109994,
          "r_table_time_ms": 232.699,
          "r_other_time_ms": 46.355,

^ Now, ANALYZE prints timing information in members named r_..._time_ms. One can see that so far, out of 30 seconds, only 232 millisecond were spent in reading the customer table. The bottleneck is elsewhere...

          "filtered": 100,
          "r_filtered": 9.085950143,
          "attached_condition": "customer.`status` = 'GOLD'"
        }
      },
      {
        "table": {
          "table_name": "orders",
          "access_type": "ref",
          "possible_keys": ["cust_id"],
          "key": "cust_id",
          "key_length": "5",
          "used_key_parts": ["cust_id"],
          "ref": ["test.customer.cust_id"],
          "r_loops": 9994,
          "rows": 1,
          "r_rows": 99.99779868,
          "r_table_time_ms": 29460.609,
          "r_other_time_ms": 986.842,

^ 29.4 seconds were spent reading the orders table (and 0.986 seconds in processing the obtained rows). Now we can see where the query is spending time.

          "filtered": 100,
          "r_filtered": 100
        }
      }
    ]
  }
}

See Also

Comments

Comments loading...
Content reproduced on this site is the property of its respective owners, and this content is not reviewed in advance by MariaDB. The views, information and opinions expressed by this content do not necessarily represent those of MariaDB or any other party.