定制的客户网格导致MySQL表扫描和文件排序又降低了性能

8

编辑：我最初在Stackoverflow上发布了这个问题，但从未得到完整的回答，我在此发布，希望能够进一步阐明为什么大规模表扫描以及任何（希望）最小的核心更改可以防止此类问题的方法。

Magento企业。1.10.1.1。客户和地址的数据集是半大型（125k +）CSR，通常位于此网格上（有时一次有25+个并发用户）。

这是在客户Grid.php Block文件中生成集合的代码片段。没什么特别的，特别是主要向集合添加属性。

$collection = Mage::getResourceModel('customer/customer_collection')
    ->addNameToSelect()
    ->addAttributeToSelect('email')
    ->addAttributeToSelect('group_id')
    ->addAttributeToSelect('prod_codes')
    ->addAttributeToSelect('last_called_date')
    ->addAttributeToSelect('time_zone')
    ->addAttributeToSelect('salesrep')
    ->addAttributeToSelect('do_not_call')
    ->addAttributeToSelect('club_member')
    ->addAttributeToSelect('call_back_date')
    ->addAttributeToSelect('marketing_code_outcome')
    ->joinAttribute('billing_postcode', 'customer_address/postcode', 'default_billing', null, 'left')
    ->joinAttribute('billing_city', 'customer_address/city', 'default_billing', null, 'left')
    ->joinAttribute('billing_telephone', 'customer_address/telephone', 'default_billing', null, 'left')
    ->joinAttribute('billing_region', 'customer_address/region', 'default_billing', null, 'left');

$this->setCollection($collection);

生成此查询，此查询行为异常，导致在客户网格中的加载时间非常长：

SELECT 
    e . *,
    _table_prefix.value AS prefix,
    _table_firstname.value AS firstname,
    _table_middlename.value AS middlename,
    _table_lastname.value AS lastname,
    _table_suffix.value AS suffix,
    CONCAT(IF(_table_prefix.value IS NOT NULL AND _table_prefix.value != '',
                CONCAT(TRIM(_table_prefix.value), ' '),
                ''),
            TRIM(_table_firstname.value),
            IF(_table_middlename.value IS NOT NULL AND _table_middlename.value != '',
                CONCAT(' ', TRIM(_table_middlename.value)),
                ''),
            ' ',
            TRIM(_table_lastname.value),
            IF(_table_suffix.value IS NOT NULL AND _table_suffix.value != '',
                CONCAT(' ', TRIM(_table_suffix.value)),
                '')) AS name,
    _table_default_billing.value AS default_billing,
    _table_billing_postcode.value AS billing_postcode,
    _table_billing_city.value AS billing_city,
    _table_billing_telephone.value AS billing_telephone,
    _table_billing_region.value AS billing_region
FROM
    customer_entity AS e
        LEFT JOIN
    customer_entity_varchar AS _table_prefix ON (_table_prefix.entity_id = e.entity_id) AND (_table_prefix.attribute_id = '4')
        LEFT JOIN
    customer_entity_varchar AS _table_firstname ON (_table_firstname.entity_id = e.entity_id) AND (_table_firstname.attribute_id = '5')
        LEFT JOIN
    customer_entity_varchar AS _table_middlename ON (_table_middlename.entity_id = e.entity_id) AND (_table_middlename.attribute_id = '6')
        LEFT JOIN
    customer_entity_varchar AS _table_lastname ON (_table_lastname.entity_id = e.entity_id) AND (_table_lastname.attribute_id = '7')
        LEFT JOIN
    customer_entity_varchar AS _table_suffix ON (_table_suffix.entity_id = e.entity_id) AND (_table_suffix.attribute_id = '8')
        LEFT JOIN
    customer_entity_int AS _table_default_billing ON (_table_default_billing.entity_id = e.entity_id) AND (_table_default_billing.attribute_id = '13')
        LEFT JOIN
    customer_address_entity_varchar AS _table_billing_postcode ON (_table_billing_postcode.entity_id = _table_default_billing.value) AND (_table_billing_postcode.attribute_id = '29')
        LEFT JOIN
    customer_address_entity_varchar AS _table_billing_city ON (_table_billing_city.entity_id = _table_default_billing.value) AND (_table_billing_city.attribute_id = '25')
        LEFT JOIN
    customer_address_entity_varchar AS _table_billing_telephone ON (_table_billing_telephone.entity_id = _table_default_billing.value) AND (_table_billing_telephone.attribute_id = '30')
        LEFT JOIN
    customer_address_entity_varchar AS _table_billing_region ON (_table_billing_region.entity_id = _table_default_billing.value) AND (_table_billing_region.attribute_id = '27')
WHERE
    (e.entity_type_id = '1')
ORDER BY CONCAT(IF(_table_prefix.value IS NOT NULL AND _table_prefix.value != '',
            CONCAT(TRIM(_table_prefix.value), ' '),
            ''),
        TRIM(_table_firstname.value),
        IF(_table_middlename.value IS NOT NULL AND _table_middlename.value != '',
            CONCAT(' ', TRIM(_table_middlename.value)),
            ''),
        ' ',
        TRIM(_table_lastname.value),
        IF(_table_suffix.value IS NOT NULL AND _table_suffix.value != '',
            CONCAT(' ', TRIM(_table_suffix.value)),
            '')) desc
LIMIT 20 OFFSET 60

在EXPLAIN上查询显示，请注意多上表E，使用临时和使用文件排序：

*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: e
         type: ref
possible_keys: IDX_ENTITY_TYPE
          key: IDX_ENTITY_TYPE
      key_len: 2
          ref: const
         rows: 55556
        Extra: Using temporary; Using filesort
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_prefix
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_VARCHAR_ATTRIBUTE,FK_CUSTOMER_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod.e.entity_id,const
         rows: 1
        Extra:
*************************** 3. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_firstname
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_VARCHAR_ATTRIBUTE,FK_CUSTOMER_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod.e.entity_id,const
         rows: 1
        Extra:
*************************** 4. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_middlename
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_VARCHAR_ATTRIBUTE,FK_CUSTOMER_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod.e.entity_id,const
         rows: 1
        Extra:
*************************** 5. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_lastname
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_VARCHAR_ATTRIBUTE,FK_CUSTOMER_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod.e.entity_id,const
         rows: 1
        Extra:
*************************** 6. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_suffix
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_VARCHAR_ATTRIBUTE,FK_CUSTOMER_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod.e.entity_id,const
         rows: 1
        Extra:
*************************** 7. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_default_billing
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_INT_ATTRIBUTE,FK_CUSTOMER_INT_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod.e.entity_id,const
         rows: 1
        Extra:
*************************** 8. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_billing_postcode
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_ADDRESS_VARCHAR_ATTRIBUTE,FK_CUSTOMER_ADDRESS_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod._table_default_billing.value,const
         rows: 1
        Extra:
*************************** 9. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_billing_city
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_ADDRESS_VARCHAR_ATTRIBUTE,FK_CUSTOMER_ADDRESS_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod._table_default_billing.value,const
         rows: 1
        Extra:
*************************** 10. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_billing_telephone
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_ADDRESS_VARCHAR_ATTRIBUTE,FK_CUSTOMER_ADDRESS_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod._table_default_billing.value,const
         rows: 1
        Extra:
*************************** 11. row ***************************
           id: 1
  select_type: SIMPLE
        table: _table_billing_region
         type: eq_ref
possible_keys: IDX_ATTRIBUTE_VALUE,FK_CUSTOMER_ADDRESS_VARCHAR_ATTRIBUTE,FK_CUSTOMER_ADDRESS_VARCHAR_ENTITY,IDX_VALUE
          key: IDX_ATTRIBUTE_VALUE
      key_len: 6
          ref: prod._table_default_billing.value,const
         rows: 1
        Extra:
11 rows in set (0.00 sec)

除了Magento本身针对1.10.1的默认索引外，没有其他索引被修改。请参见此处的1.5.1（CE）结构：http : //www.magereverse.com/index/magento-sql-structure/version/1- 5-1-0

这是称为AS e的别名表。在扫描中：

CREATE TABLE `customer_entity` (
    `entity_id` INT(10) UNSIGNED NOT NULL AUTO_INCREMENT,
    `entity_type_id` SMALLINT(8) UNSIGNED NOT NULL DEFAULT '0',
    `attribute_set_id` SMALLINT(5) UNSIGNED NOT NULL DEFAULT '0',
    `website_id` SMALLINT(5) UNSIGNED NULL DEFAULT NULL,
    `email` VARCHAR(255) NOT NULL DEFAULT '',
    `group_id` SMALLINT(3) UNSIGNED NOT NULL DEFAULT '0',
    `increment_id` VARCHAR(50) NOT NULL DEFAULT '',
    `store_id` SMALLINT(5) UNSIGNED NULL DEFAULT '0',
    `created_at` DATETIME NOT NULL DEFAULT '0000-00-00 00:00:00',
    `updated_at` DATETIME NOT NULL DEFAULT '0000-00-00 00:00:00',
    `is_active` TINYINT(1) UNSIGNED NOT NULL DEFAULT '1',
    PRIMARY KEY (`entity_id`),
    INDEX `FK_CUSTOMER_ENTITY_STORE` (`store_id`),
    INDEX `IDX_ENTITY_TYPE` (`entity_type_id`),
    INDEX `IDX_AUTH` (`email`, `website_id`),
    INDEX `FK_CUSTOMER_WEBSITE` (`website_id`),
    CONSTRAINT `FK_CUSTOMER_ENTITY_STORE` FOREIGN KEY (`store_id`) REFERENCES `core_store` (`store_id`) ON UPDATE CASCADE ON DELETE SET NULL,
    CONSTRAINT `FK_CUSTOMER_WEBSITE` FOREIGN KEY (`website_id`) REFERENCES `core_website` (`website_id`) ON UPDATE CASCADE ON DELETE SET NULL
)

因此，问题是如何使该查询更好地执行，而又不导致临时表被创建和扫描。

我不确定我可以索引哪些内容来改善此查询性能，并且我不想过多地研究修改Magento的ORM。

mysql  grid  performance  grid-serlization  mysql-indexes 
— B00MER
source

Answers:

2

困境的答案不在于定制，这些定制就可以了。查询中有问题的部分是ORDER BY，并且仅当通过admin中的“名称”列对网格进行排序时，才会抬起丑陋的头：

ORDER BY CONCAT(IF(_table_prefix.value IS NOT NULL AND _table_prefix.value != '',
        CONCAT(TRIM(_table_prefix.value), ' '),
        ''),
    TRIM(_table_firstname.value),
    IF(_table_middlename.value IS NOT NULL AND _table_middlename.value != '',
        CONCAT(' ', TRIM(_table_middlename.value)),
        ''),
    ' ',
    TRIM(_table_lastname.value),
    IF(_table_suffix.value IS NOT NULL AND _table_suffix.value != '',
        CONCAT(' ', TRIM(_table_suffix.value)),
        '')) DESC

需要使用此ORDER BY子句按“名称”列进行排序，因为“名称”列是客户的前缀，名字，中间名称和姓氏的串联。要组合信息并将其用于排序，必须使用tmp表，因为要串联在一起的数据源自查询中的JOIN。这就是Magento内核处理这种排序的方式，因为它只是应用WHERE子句中使用的相同代码来应用排序。

将整个查询包装为子选择将减少查询时间，因为它将不再使用临时表，但仍将使用文件排序。

因此，在这里可以看到两个“解决方案”：

  1. 在保存时填充的客户实体上创建一个full_name属性。然后，可以在网格上使用此预先填充的full_name属性，而无需进行任何形式的串联，并且可以消除对临时表和文件排序的需求。

  2. 当时间成为问题时，根本就不按“名称”列排序。如果您需要按名称排序，这显然无济于事，但要记住一点，直到您有机会构建上面＃1中的更改​​为止。:)

— 大卫·阿尔格
source
感谢您对ORDER BY条款的回答和澄清，以及first，lastname，prefix的加入。我最初的计划是简单地打平客户，地址表并重写一些逻辑，这两个其他选项听起来更快捷，更容易实现。再次感谢。
— B00MER
By using our site, you acknowledge that you have read and understand our Cookie Policy and Privacy Policy.
Licensed under cc by-sa 3.0 with attribution required.