优化MySQL CREATE TABLE查询

Question

I have two tables I am trying to join in a third query and it seems to be taking far too long.

Here is the syntax I am using

CREATE TABLE active_users
(PRIMARY KEY ix_all (platform_id, login_year, login_month, person_id))
SELECT platform_id
    , YEAR(my_timestamp) AS login_year
    , MONTH(my_timestamp) AS login_month
    , person_id
    , COUNT(*) AS logins
FROM
    my_login_table
GROUP BY 1,2,3,4;

CREATE TABLE active_alerts
(PRIMARY KEY ix_all (platform_id, alert_year, alert_month, person_id))
SELECT platform_id
    , YEAR(alert_datetime) AS alert_year
    , MONTH(alert_datetime) AS alert_month
    , person_id
    , COUNT(*) AS alerts
FROM 
    my_alert_table
GROUP BY 1,2,3,4;

CREATE TABLE all_data
(PRIMARY KEY ix_all (platform_id, theYear, theMonth, person_id))
SELECT a.platform_id
    , a.login_year AS theyear
    , a.login_month AS themonth
    , a.person_id
    , IFNULL(a.logins,0) AS logins
    , IFNULL(b.alerts,0) AS job_alerts
FROM
    active_users a
LEFT OUTER JOIN
    active_alerts b
        ON a.platform_id = b.platform_id
        AND a.login_year = b.alert_year
        AND a.login_month = b.alert_month
        AND a.person_id = b.person_id;

The first table (logins) returns about half a million rows and takes less than 1 minute, the second table (alerts) returns about 200k rows and takes less than 1 minute.

If I run just the SELECT part of the third statement it runs in a few seconds, however as soon as I run it with the CREATE TABLE syntax it takes more than 30 minutes.

I have tried different types of indexes than a primary key, such as UNIQUE or INDEX as well as no key at all, but that doesn't seem to make much difference.

Is there something I can do to speed up the creation / insertion of this table?

EDIT: Here is the output of the show create table statements

CREATE TABLE `active_users` (
  `platform_id` int(11) NOT NULL,
  `login_year` int(4) DEFAULT NULL,
  `login_month` int(2) DEFAULT NULL,
  `person_id` varchar(40) NOT NULL,
  `logins` bigint(21) NOT NULL DEFAULT '0',
  KEY `ix_all` (`platform_id`,`login_year`,`login_month`,`person_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci

CREATE TABLE `alerts` (
  `platform_id` int(11) NOT NULL,
  `alert_year` int(4) DEFAULT NULL,
  `alert_month` int(2) DEFAULT NULL,
  `person_id` char(36) CHARACTER SET ascii COLLATE ascii_bin NOT NULL,
  `alerts` bigint(21) NOT NULL DEFAULT '0',
  KEY `ix_all` (`platform_id`,`alert_year`,`alert_month`,`person_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci

and the output of the EXPLAIN

id  select_type table   partitions  type    possible_keys   key key_len ref rows    filtered    Extra

1   SIMPLE  a   (null)  ALL (null)  (null)  (null)  (null)  503504  100 (null)

1   SIMPLE  b   (null)  ALL ix_all  (null)  (null)  (null)  220187  100 Using where; Using join buffer (Block Nested Loop)

Answer 1

It's a bit of a hack but I figured out how to get it to run much faster.

I added a primary key to the third table on platform, year, month, person

I inserted the intersect data using an inner join, then insert ignore the left table plus a zero for alerts in a separate statement.