简体繁体 English

Mysql全局变量'innodb_file_format'值在主服务器和从服务器上有所不同

[英]Mysql Global variable 'innodb_file_format' value differs in master and slave

原文 2017-05-14 22:51:30 4 1 mysql/ file/ format/ innodb

While debugging the slave is lagging behind master, I have observed global variable innodb_file_format differs in slave and master machine. 虽然调试从机是落后于主机，但我观察到从机和主机的全局变量innodb_file_format不同。 Master is set to Barracuda and slave is set to Antelope . Master被设置为Barracuda ，奴隶被设置为Antelope 。 Does it have the impact on slave lagging? 它对奴隶滞后有影响吗？

1 个解决方案

5000 seconds is 83 minutes. 5000秒是83分钟。 That is more than 22 minutes, so something else is going on. 那是超过22分钟，所以还有其他事情正在发生。

The file_format probably would not explain the difference. file_format 可能无法解释差异。 What is binlog_format ? 什么是binlog_format ？ (That probably does not matter either, just want to get it in play.) （这也许并不重要，只是想让它发挥作用。）

Let's see the slowest queries, together with SHOW CREATE TABLE for the relevant table(s). 让我们看一下最慢的查询，以及相关表的SHOW CREATE TABLE 。 We can probably suggest how to speed some of them up. 我们可以建议如何加速其中一些。

innofb_flush_log_at_trx_commit may have a big impact IO performance. innofb_flush_log_at_trx_commit 可能会对IO性能产生很大影响。 Do you have HDD or SSD? 你有硬盘还是SSD？ How many transactions were run in those 22 minutes? 在这22分钟内运行了多少笔交易？ =2 is faster, but less safe -- generally good for a Slave. =2更快，但安全性更低 - 通常对奴隶有利。 =1 is slower, but safer -- generally good for a Master. =1更慢，但更安全 - 通常对主人有利。 I suggest you got them backwards. 我建议你让他们倒退。

Usually a slave lag is caused by transactions are run in parallel (from different connections) on the Master. 通常，从属延迟是由主服务器上并行（来自不同连接）的事务引起的。 Is that the case? 是这样的吗？ And the Slave is single threaded. 奴隶是单线程的。 What version of MySQL? 什么版本的MySQL？ Only newer versions even have the option of parallelism. 只有较新的版本甚至可以选择并行。