Azure autoscaling - how it works from my observation



This is my observation with scaling out app service plan in an ASE environment.

How long does it takes to scale?
30 to 40 min to scale up one instance. Your service plan status will be pending.

Where do you view your metrics / logs?
Go to app insights in your function app or apps. For some reason it is not updated

What metric to choose?

That depends. For my case, I have to scale function app. So i set it scale if function app execution count is greater than 8. I noticed that regardless of how many instance you already running, for example you have 3 instance, as long as function app execution is 8 then it will scale up to 4 instance.




It is not like function app execution unit shared by current 3 instances. If you have 9 function app, then you get a new instance.

Off course it differ depending on metrics types.

Event hubs

As i push more load to event hubs, partitioning is pretty much fixed. If you have 2 VM, then those functionapp will be processing the same partition. Process wpw3 is always one.

If loads becomes massive, then you can see event hub start to send in more data in a batch. In my case, when the server is not swamped, then event messages delivered are 1. When servers are under heavy loads, messages tend to be more.


Comments

Popular posts from this blog

Solving Sonarqube :- Project was never analyzed. A regular analysis is required before a branch analysis

PySpark - Working with JDBC Sqlite database