SQL数据库DTU缩放速度慢？ [英] SQL Database DTU scaling slow?

问题描述

我们几乎每天都在扩展主要的生产数据库。我们从S2或S3开始，然后在早上扩展到P1和P4之间的某个位置，然后在一天结束时再次降低。我们已经这么做了多年。这个缩放*总是*大约15-40分钟需要
。然而，今天这对我们来说已经明显放缓了。而不是半小时，我们看到它需要超过2小时（目前升级现在即将达到4小时）。

给出了什么？

解决方案

您的数据库有多大？

我做了很多扩展我们的开发环境，每天可以放五到十次，偶尔我可以扩展生产 - 但是我讨厌这样做，因为我们是一个24x7的操作，并且扩展几乎总是会导致活动交易。

我们的数据库只有大约20gb，*通常*它在5分钟左右完成。  但是现在可能需要更长时间。  微软在这个问题上唯一的说法就是没有SLA，但如果*是* SLA则可以保证在大约6小时内完成
。

（是的，我说了几个小时）

现在我认为这是比我的更大的数据库，并且似乎需要额外的时间从P到S或再次返回。

所以我可以告诉你的是它并不漂亮，但微软并不认为它不受限制。

如果您有支持计划，您仍然可以点击它并分享你对他们的担忧，但一般来说他们无能为力，特别是短短六个小时，但告诉你如何用以下的方式监控它：

选择*来自sys.dm_operation_status; 

（必须从主人运行）

Josh

ps - 当然还要检查服务中断 - 除了门户网站的小服务健康小部件似乎在周末已被弃用。

We scale our main production database on an almost-daily basis. We start out at either S2 or S3 and scale up to somewhere between P1 and P4 in the mornings, then back down again at the end of the day. We've done this for years. This scaling *always* takes around 15-40 minutes. However, this has slowed down significantly today for us. Instead of ~half an hour, we're seeing it take over 2 hours (current upgrade is now about to hit 4 hours).

What gives?

解决方案

How large is  your database?

I've done a lot of scaling our dev environments, can be five or ten times a day, and now and then I scale production - but hate doing so because we're a 24x7 operation and scaling almost always kills active transactions.

Our database is only about 20gb, and *usually* it gets done within five minutes or so.  But now and then it can take much longer.  Microsoft's only word on the issue is that there is no SLA, but if there *were* an SLA it might guarantee completion in about six hours.

(yes, I said hours)

Now I assume that's for much larger databases than mine, and it seems to take extra time to cross from P to S or back again.

So all I can tell you is it's not pretty, but Microsoft does not consider it out of bounds.

If you have a support plan you can still click on it and share your concerns with them, but generally there's nothing they can do, especially short of six hours, but tell you how to monitor it with something like:

select * from sys.dm_operation_status;

(must run from master)

Josh

ps - and of course check for service outages - except the portal's little service health widget seems to have been deprecated over the weekend.

这篇关于SQL数据库DTU缩放速度慢？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！