Clever Geek Handbook

The load in some instances of the Google App Engine application is very low in others. What for?

As you can see in the next snapshot, the load is huge in some dynamic instances (over 20K requests), while in others it is very small.

enter image description here

Why is this happening? Shouldn't the load be distributed evenly?

+3


source to share


2 answers


This is what I got from a Google App Engine expert:

Request Scheduling App Engine uses several heuristics for routing requests to application instances. When the QPS is low, it remains in affinity scheduling mode and redirects most requests to the instances that most recently responded to the health check and processed the requests successfully. This explains why you see this change in the number of requests for each instance. When you add traffic to your application, the load should be equal to all instances.

I also asked what was GAE's policy for closing instances. I can see that many of them are ready even if they don't receive any request



Dynamic instances that don't serve requests end up being garbage collected. However, you only receive an invoice for 15 extra minutes after they serve the last request. For more information on billing, see this document.

https://cloud.google.com/appengine/kb/billing#different_on_demand_instance_resident

+1


source


If the load is balanced across active dynamic instances, they will rarely become idle (only when the entire application traffic drops to near zero), so it would be difficult to dynamically close them down.

More details here:



+2


source






More articles:


All Articles