A scaling policy issue

Hello

I set up 2 concurrent processes in a fleet like the picture below.

We needed 2 more dedicated servers. So I turned on a scaling policy like the picture below.

A couple of minutes after the scaling policy was turned on, we encountered this strange error like the picture below.


After the interruption, our matchmaking requests failed every time . So I just turned off the scaling policy and turn it on again and then we could submit our matchmaking requests to FlexMatch.

Why did it happen?

Just in case, I left the information.
FleetId : fleet-a520ea76-5989-474b-9915-8ca6b46dea2c
First GameSessionId : arn:aws:gamelift:us-west-2::gamesession/fleet-a520ea76-5989-474b-9915-8ca6b46dea2c/325a866e-6ba0-40c6-9a3e-aa2c776d567f
Second GameSessionId : arn:aws:gamelift:us-west-2::gamesession/fleet-a520ea76-5989-474b-9915-8ca6b46dea2c/83fcc5ad-e974-4e49-9f3c-52838592b3e7

Hi there,

Thanks for including detailed information. We will take a look at what happened to your fleet and get back to you. Before that I think you want to make sure that you have enough server processes configured per instance and have enough active server processes for each instance when placing your game sessions.

Note that you mentioned you needed 2 more instances but the maximum amount instances you can have with your scaling policy is 2.

I don’t know the exact terms in GameLit but a fleet activates an EC2 instance to launch dedicated servers. If a scaling policy is turned on, FlexMatch activates more EC2 instances if needed. This point is important. From what I understand, “Concurrent processes” in the picture below, means the maximum number of processes in each EC2 instance.

Before the scaling policy was turned on, only an EC2 instance that could launch 2 dedicated servers at the same time was activated. However, we needed 2 more dedicated servers. So the scaling policy was turned on and then one more EC2 instance was activated. So totally, 4 dedicated servers could be activated at that time. Bottom line is that 2 EC2 instances were activated. Each EC2 instance could launch 2 dedicated servers at the same time. (2 EC2 instances * 2 dedicated servers)

In this scenario, I’m sure that we had enough active server processes for each instance.

Thanks for the quick response

Hi there,

I think what happened was that you did not turn on instance protection on your fleet. And the host (i-048aa68db11a64b00) that were hosting the game sessions were terminated after you enable autoscaling due to low server utilization rate. If you turn on instance protection here it will prevent game instances being terminated by AutoScaling if they have active game sessions on them.

https://docs.amazonaws.cn/en_us/gamelift/latest/apireference/API_CreateFleet.html#gamelift-CreateFleet-request-NewGameSessionProtectionPolicy

Thank you for giving me the important information.