posit labs
posit labs

Reputation: 9431

Deploy >20 instances on Appengine

I need to deploy a bunch of instances for bursts of traffic at the same time every week (auto scaling can't handle these extreme bursts), but I seem to be stuck at a maximum of 20 instances.

These docs say that 20 is the cap for free apps, but I have a billing account connected. I even have the app running in us-central, which should be able to go up to 200 instances.

https://cloud.google.com/appengine/docs/standard/python/an-overview-of-app-engine

Re: David Neale - The app.yaml

runtime: nodejs8 
instance_class: B4 
manual_scaling:
  instances: 21

Re: Dan Cornilescu - The goal is to support up to 50k concurrent users of a web app. The burst will coincide with a weekly event, once a week, for about 3 hours. We are expecting the biggest burst at the start of the event. We've been advised against using anything that requires lead time to scale up since the spikes are so extreme.

I haven't confirmed this, but the main reason we need more instances isn't actually the processor load, but rather the number of ports available to appengine instances for incoming requests (I can't find docs on max connections). There is 3rd party API we need to use, and one of the requests is taking 250ms. I think all of the available ports are being used up just waiting on this call to complete.

Upvotes: 3

Views: 94

Answers (1)

Ying Li
Ying Li

Reputation: 2519

If your app has billing account attached and you don't have a daily budget that's reached, then you shouldn't have any problem getting 200 manually scaled instances in us-central since the limit is in fact 200.

You could also simply split your traffic between different versions since every version has their own scaling limit. So even if they are for some reason stuck at 20 each, you can setup more versions and split traffic across them. You must have billing enabled and working of course.

Upvotes: 2

Related Questions