I am having a few issues with the AWS EC2 autoscaling feature. It is kinda working but exhibiting some undesirable results.
The default EC2 location specified in my settings.ini file is us-east-1. Whatever I have set for the max instance field is how many agents are being spun up. So, initially, I had max=4 and 4 agents would spin up even though I dont think I have active tests queued up…? I have gone to /work/jobs/us-east-1 and cleared that folder and nothing changes. I have noticed that the folder repopulates so I am not sure if it is pulling from a different location.
Also, I have noticed that the instances in the default location are terminating (seemingly because of idle time, set to 8 minutes, currently) and then concurrently spinning up a new agent instance. I have only seen this in the default test location.
The 2 other locations that I have loaded, for now, are spinning up and spinning down test agents as I input tests. I noticed that the first byte is bad maybe because the server is in us-east-1 and I have agents in ap-southeast-2 (Sydney).
Would it make sense to have a server in each desired location that autoscales their own agents for more timely results?
Ultimately, I will be operating this on the API FWIW.
Also, is the scalefactor as straightforward as I am reading it? The value is the number of tests in the queue before a new agent is spun up? For example, if the value is set to 25, and I have a max agent value of 2, and I have 75 tests, it will only spin up 2 agents?
Thanks in advance for any help.