Sometime Fleet is slowly activated

I pushed a dedicated server for Linux to GameLift. The dedicated server normally takes 15mins to be activated but it took about 40mins. Sometimes it happens but why does it happen?

If you see this happening repeatedly, I would first suggest opening a ticket through AWS support for a faster response.

To activate a fleet, GameLift needs to do a lot of things: get a new EC2 instance, download your game on to it, install it and wait for it to go active so a lot can slow this down:

  • Provisioning an EC2 instance always takes time regardless of the health of the system. Anecdotally Windows instance always seem to take longer than Linux instances.
    • EC2 can at times provide bad instances which can fail in a bunch of ways (unreachable or not responsive, can’t be placed in VPCs etc) or temporary lack of capacity of very specific instances types in a region/availability zone.
  • Your binaries need to be grabbed and downloaded and installed which relies on storage or network systems not having transient issues and high throughput.
  • GameLift builds on top of other AWS services and so they also need to be healthy and performant.

You can always check AWS service health here: to see if there are any significant issues in a region/service.

GameLift is designed to be reliable and fault tolerant but occasionally may need to retry provisioning steps.

In this case there was a clearly a delay in getting that initial instance and placing your build up on it. That was probably frustrating to wait that long for a fleet.

I’ll alert the GameLift service team to this so they can check everything is working as expected.

BTW what region was this fleet in?

This fleet is in US West (Oregon), us-west-2.

Thank you

Hi there,

I took at look at your fleet and the events around your fleet creation. There isn’t anything out of ordinary. Sometimes it will take a while for our system to process your fleet if there are too many fleets pending creation. So it took around 16 minutes to get to your fleet in this case. The system works in this way to avoid causing internal issues when there is a spike of fleet creation globally.

Thanks you very much.
I got it. It’s an ordinary situation.