为什么 MySQL 在此查询中忽略索引？

Question

我有以下表格：

CREATE TABLE `sms` (
  `sms_id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `sms_datetime` datetime DEFAULT NULL,
  `sms_number` varchar(40) NOT NULL,
  `sms_text` text,
  `sms_status` int(3) DEFAULT '0',
  `sms_last_tm` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
  `sms_csinteve_id` int(11) unsigned NOT NULL DEFAULT '0',
  `sms_handler_user_id` int(11) unsigned NOT NULL DEFAULT '0',
  PRIMARY KEY (`sms_id`),
  KEY `sms_number` (`sms_number`),
  KEY `sms_last_tm` (`sms_last_tm`),
  KEY `sms_csinteve_id` (`sms_csinteve_id`),
  KEY `sms_handler_user_id` (`sms_handler_user_id`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8;

sms 的行数比 2,000,000 多。

还有下表：

CREATE TABLE `customer_service_interaction_events` (
  `csinteve_id` int(11) unsigned NOT NULL AUTO_INCREMENT,
  `csinteve_interaction_id` int(11) unsigned NOT NULL DEFAULT '0',
  `csinteve_datetime` datetime DEFAULT NULL,
  PRIMARY KEY (`csinteve_id`),
  KEY `csinteve_interaction_id` (`csinteve_interaction_id`),
  KEY `csinteve_datetime` (`csinteve_datetime`)
) ENGINE=InnoDB AUTO_INCREMENT=21153 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;

20,000 行。

然后这个查询：

EXPLAIN SELECT COUNT(csinteve_id) as interactions_number
FROM customer_service_interaction_events
JOIN sms
    ON csinteve_id = sms_csinteve_id
WHERE csinteve_interaction_id = 3085
AND sms_handler_user_id = 0;

说：

{
    "data":
    [
        {
            "id": 1,
            "select_type": "SIMPLE",
            "table": "customer_service_interaction_events",
            "partitions": null,
            "type": "ref",
            "possible_keys": "PRIMARY,csinteve_interaction_id",
            "key": "csinteve_interaction_id",
            "key_len": "4",
            "ref": "const",
            "rows": 1,
            "filtered": 100,
            "Extra": "Using index"
        },
        {
            "id": 1,
            "select_type": "SIMPLE",
            "table": "sms",
            "partitions": null,
            "type": "ALL",
            "possible_keys": "sms_csinteve_id,sms_handler_user_id",
            "key": null,
            "key_len": null,
            "ref": null,
            "rows": 2083577,
            "filtered": 99.4,
            "Extra": "Using where; Using join buffer (Block Nested Loop)"
        }
    ]
}

说对于sms 表MySQL 不使用任何索引（类型ALL）。 而sms 表在与customer_service_interaction_events 连接的sms_csinteve_id 列上有一个索引，而customer_service_interaction_events 表在csinteve_interaction_id 上有一个索引。

查询的目的是返回未处理的短信数量（sms），每条短信绑定一个客服事件（customer_service_interaction_events），多个客服事件有相同的交互id（customer_service_interaction_events.csinteve_interaction_id） , 在示例中为 3085)。

谢谢。

编辑：

我尝试添加 Akina 的索引：

CREATE INDEX idx ON customer_service_interaction_events (csinteve_interaction_id, csinteve_id);
CREATE INDEX idx ON sms (sms_handler_user_id, sms_csinteve_id);

但是解释：

EXPLAIN SELECT COUNT(csinteve_id) as interactions_number
FROM customer_service_interaction_events
JOIN sms
    ON csinteve_id = sms_csinteve_id
WHERE csinteve_interaction_id = 3085
AND sms_handler_user_id = 0;

仍然为sms 输出类型ALL：

{
    "data":
    [
        {
            "id": 1,
            "select_type": "SIMPLE",
            "table": "customer_service_interaction_events",
            "partitions": null,
            "type": "ref",
            "possible_keys": "PRIMARY,csinteve_interaction_id,csinteve_interaction_id__csinteve_id",
            "key": "csinteve_interaction_id",
            "key_len": "4",
            "ref": "const",
            "rows": 1,
            "filtered": 100,
            "Extra": "Using index"
        },
        {
            "id": 1,
            "select_type": "SIMPLE",
            "table": "sms",
            "partitions": null,
            "type": "ref",
            "possible_keys": "sms_csinteve_id,sms_handler_user_id,sms_csinteve_id_2,sms_handler_user_id_2,sms_csinteve_id_3,sms_handler_user_id_3,sms_csinteve_id_4,sms_handler_user_id_4,sms_handler_user_id__sms_csinteve_id",
            "key": "sms_handler_user_id__sms_csinteve_id",
            "key_len": "8",
            "ref": "const,local_bluelinks.customer_service_interaction_events.csinteve_id",
            "rows": 2083577,
            "filtered": 100,
            "Extra": "Using index"
        }
    ]
}

Answer 1

测试一下：

CREATE INDEX idx ON customer_service_interaction_events (csinteve_interaction_id, csinteve_id);
CREATE INDEX idx ON sms (sms_handler_user_id, sms_csinteve_id);