Question

我有一个带有MySQL db（所有innodb表）的Drupal站点，每隔一段时间就会“崩溃”（读取：运行缓慢）。服务器托管基于云，当数据库移至新的容器时会发生“崩溃”。联接大约10个表并返回一行的查询在移动后开始运行非常缓慢（1分钟），并且一切都停止了。通常，查询执行速度非常快（少于.1秒），这不是问题。

已确定此移动会导致“崩溃”，因为MySQL服务器丢失了其“分析”信息-存储在内存中，并且在移动数据库时也会丢失。

移动后当前无法自动运行分析。托管公司建议重新编写查询，以便不必进行分析。我想问社区这是否有意义，以及我将如何重写它。

以下是从OP的pastebin复制的信息：

查询时间：0.052842锁定时间：0.000530已发送的行数：1已检查的行数：61031

SELECT node__field_last_sl_play.field_last_sl_play_value AS node__field_last_sl_play_field_last_sl_play_value, node_field_data.nid AS nid, node_field_data_node__field_album.nid AS node_field_data_node__field_album_nid, node_field_data_node__field_artist.nid AS node_field_data_node__field_artist_nid, node_field_data_node__field_event.nid AS node_field_data_node__field_event_nid, votingapi_result_node_field_data.id AS votingapi_result_node_field_data_id, node_field_data_node__field_lifestyle.nid AS node_field_data_node__field_lifestyle_nid
FROM
node_field_data node_field_data
LEFT JOIN node__field_album node__field_album ON node_field_data.nid = node__field_album.entity_id AND node__field_album.deleted = '0' AND (node__field_album.langcode = node_field_data.langcode OR node__field_album.bundle = 'track')
LEFT JOIN node_field_data node_field_data_node__field_album ON node__field_album.field_album_target_id = node_field_data_node__field_album.nid
LEFT JOIN node__field_artist node_field_data_node__field_album__node__field_artist ON node_field_data_node__field_album.nid = node_field_data_node__field_album__node__field_artist.entity_id AND node_field_data_node__field_album__node__field_artist.deleted = '0' AND (node_field_data_node__field_album__node__field_artist.langcode = node_field_data_node__field_album.langcode OR node_field_data_node__field_album__node__field_artist.bundle = 'album')
LEFT JOIN node_field_data node_field_data_node__field_artist ON node_field_data_node__field_album__node__field_artist.field_artist_target_id = node_field_data_node__field_artist.nid
LEFT JOIN node__field_event node__field_event ON node_field_data.nid = node__field_event.entity_id AND node__field_event.deleted = '0' AND (node__field_event.langcode = node_field_data.langcode OR node__field_event.bundle = 'track')
LEFT JOIN node_field_data node_field_data_node__field_event ON node__field_event.field_event_target_id = node_field_data_node__field_event.nid
LEFT JOIN votingapi_result votingapi_result_node_field_data ON node_field_data.nid = votingapi_result_node_field_data.entity_id AND (votingapi_result_node_field_data.entity_type = 'node' AND votingapi_result_node_field_data.function = 'vote_sum' AND votingapi_result_node_field_data.type = 'vote')
LEFT JOIN node__field_lifestyle node__field_lifestyle ON node_field_data.nid = node__field_lifestyle.entity_id AND node__field_lifestyle.deleted = '0'
LEFT JOIN node_field_data node_field_data_node__field_lifestyle ON node__field_lifestyle.field_lifestyle_target_id = node_field_data_node__field_lifestyle.nid
LEFT JOIN node__field_last_sl_play node__field_last_sl_play ON node_field_data.nid = node__field_last_sl_play.entity_id AND node__field_last_sl_play.deleted = '0'
WHERE (node_field_data.status = '1') AND (node_field_data.type IN ('track'))
ORDER BY node__field_last_sl_play_field_last_sl_play_value DESC
LIMIT 1 OFFSET 0;

此查询的EXPLAIN报告：

EXPLAIN
1   SIMPLE  node_field_data index_merge node_field__type__target_id,node__status_type   node_field__type__target_id,node__status_type   34,35   
    NULL
    2288    Using intersect(node_field__type__target_id,node__status_type); Using where; Using index; Using temporary; Using filesort   
1   SIMPLE  node__field_album   ref PRIMARY,bundle  PRIMARY 5   dbname.node_field_data.nid,const    1   Using where 
1   SIMPLE  node_field_data_node__field_album   ref PRIMARY,node__id__default_langcode__langcode    PRIMARY 4   dbname.node__field_album.field_album_target_id  1   Using where; Using index    
1   SIMPLE  node_field_data_node__field_album__node__field_artist   ref PRIMARY,bundle  PRIMARY 5   dbname.node_field_data_node__field_album.nid,const  1   Using where 
1   SIMPLE  node_field_data_node__field_artist  ref PRIMARY,node__id__default_langcode__langcode    PRIMARY 4   dbname.node_field_data_node__field_album__node__field_artist.field_artist_target_id 1   Using where; Using index    
1   SIMPLE  node__field_event   ref PRIMARY,bundle  PRIMARY 5   dbname.node_field_data.nid,const    1   Using where 
1   SIMPLE  node_field_data_node__field_event   ref PRIMARY,node__id__default_langcode__langcode    PRIMARY 4   dbname.node__field_event.field_event_target_id  1   Using where; Using index    
1   SIMPLE  votingapi_result_node_field_data    ref vote_result_field__type__target_id,vote_result_field__entity_id__target_id  vote_result_field__entity_id__target_id 5   dbname.node_field_data.nid  1   Using where 
1   SIMPLE  node__field_lifestyle   ref PRIMARY PRIMARY 5   dbname.node_field_data.nid,const    1       
1   SIMPLE  node_field_data_node__field_lifestyle   ref PRIMARY,node__id__default_langcode__langcode    PRIMARY 4   dbname.node__field_lifestyle.field_lifestyle_target_id  1   Using where; Using index    
1   SIMPLE  node__field_last_sl_play    ref PRIMARY PRIMARY 5   dbname.node_field_data.nid,const    1

表定义：

CREATE TABLE `node_field_data` (
 `nid` int(10) unsigned NOT NULL,
 `vid` int(10) unsigned NOT NULL,
 `type` varchar(32) CHARACTER SET ascii NOT NULL COMMENT 'The ID of the target entity.',
 `langcode` varchar(12) CHARACTER SET ascii NOT NULL,
 `title` varchar(255) NOT NULL,
 `uid` int(10) unsigned NOT NULL COMMENT 'The ID of the target entity.',
 `status` tinyint(4) NOT NULL,
 `created` int(11) NOT NULL,
 `changed` int(11) NOT NULL,
 `promote` tinyint(4) NOT NULL,
 `sticky` tinyint(4) NOT NULL,
 `revision_translation_affected` tinyint(4) DEFAULT NULL,
 `default_langcode` tinyint(4) NOT NULL,
 `rh_action` varchar(255) DEFAULT NULL,
 `rh_redirect` varchar(255) DEFAULT NULL,
 `rh_redirect_response` int(11) DEFAULT NULL,
 PRIMARY KEY (`nid`,`langcode`),
 KEY `node__id__default_langcode__langcode` (`nid`,`default_langcode`,`langcode`),
 KEY `node__vid` (`vid`),
 KEY `node_field__type__target_id` (`type`),
 KEY `node_field__created` (`created`),
 KEY `node_field__changed` (`changed`),
 KEY `node__frontpage` (`promote`,`status`,`sticky`,`created`),
 KEY `node__status_type` (`status`,`type`,`nid`),
 KEY `node__title_type` (`title`(191),`type`(4)),
 KEY `node_field__uid__target_id` (`uid`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='The data table for node entities.'

Answer 1

codesmith，您的查询没有问题。当表具有合理的索引而不是1分钟时，它将在.1秒内运行。由于您和您的托管公司都知道向另一个容器移动需要进行分析才能完成该过程，所以为什么这不是他们的过程的一部分？ “ FWIW这是一家著名的托管公司，使用Google进行容器化。”尝试说服托管公司加倍努力，并分析每张桌子-在他们“移动”之后，以避免在随后几天（或几周）出现“异常长时间运行的查询”时避免此发现。如果时间允许，我仍然想和您一起Skype TALK。

是否可以优化查询，从而不必运行分析？

1 个答案: