每10分钟Azure数据库池出现意外峰值 [英] Unexpected spikes in Azure database pool every 10 minutes

问题描述

您好

我注意到我们的Azure SQL池每10分钟就会显示数据IO中的峰值（参见屏幕截图）。我在池中有大约140个数据库。当我查看单个数据库时，我看不到相同的峰值。

I've noticed our Azure SQL pool shows spikes in Data IO every 10 minutes (see screenshot). I have about 140 databases in the pool. When I look at individual databases I don't see the same spikes.

如何确定哪些数据库（如果有）负责。是否有任何Azure系统进程可以执行此操作（日志备份）？ 

How can I determine which databases, if any, are responsible. Are there any Azure system processes which would do this (log backups)? 

非常感谢

推荐答案

美好的一天<？！？>

Good day <?!?>

当您将所有数据库组合在一起时，每个数据库中的小峰值可能会成为一个巨大的峰值。

A small spike in each database might become a huge spike when you combine all the databases together.

让我们考虑一个理论案例：你有60个数据库，并且每个小时（同时）都有2个DTU的峰值，但是在其他59分钟内这些峰值之间每个数据库使用0 DTU并且他每个数据库在另一个小时内还有另一个误差
。

Let's think about a theoretical case: you have 60 databases and all have spike of 2 DTU every round hour (at the same time), but between these spikes during the other 59 minutes each database use 0 DTU and he each DB has another spite in a different minute in the hour.

如果你分别查看每个数据库，那么就没有尖峰。它使用1-2 DTU，但是现在检查所有数据库的DTU的SUM ...在圆形小时之间，值将正好是2 DTU，因为在每分钟不同的数据库有2个DTU的峰值和
休息使用0 DTU，但是圆形小时你获得60 DTU，因为此时所有数据库都有峰值......

If you look at each DB separately then there is no spike. it uses 1-2 DTU, but now check the SUM of the DTU for all databases... between the round hour the values will be exactly 2 DTU since in each minute a different database has a spike of 2 DTU amd the rest use 0 DTU, but the round hour you get 60 DTU since all databases has spike at this time...

下一步是什么？ 监控

1。检查一半数据库的行为，然后检查另一半

1. Check the behavior of half of the database together, and next check the other half

如果在两个组中都出现峰值，则问题与特定的单个数据库无关（您可以再试一次）通过使用4组确认没有）

If in both groups you get a spike then the issue is not related to specific single database (you can try again to confirm by using 4 groups no)

如果只有一个组有一个峰值，那么问题与该组有关，您可以再次拆分并检查数据库的每一半

If only one group has a spike then the issue related to this group and you can once more split and check each half of the database

2。很明显，这是一致的行为，所以有一些事情发生。使用扩展事件来监视在此期间数据库级别上发生的事情。 

2. It is pretty clear that this is consistent behavior so there is something there that happen. Use extended events to monitor what happens on the databases level during that time. 

可能与Azure的维护有关的问题，因为资源太低。尝试使用更高的定价等级并为我们带来行为的结果

Maybe the issue related to maintenance by the Azure, since the resources are too low. try to use a higher pricing tier and bring us the result of the behavior

这篇关于每10分钟Azure数据库池出现意外峰值的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！