Selenium Grid Auto Scaling
Your selenium grid is using auto scaling and reacts to pending environment requests and will launch selenium grid nodes to fulfill the demand. Auto scaling requires all 3 test environment parameters platform, browser and browser version to be specified in each request in order to launch grid nodes to process test requests. Incorrect requests are ignored by the auto scaling functionality and the test will time out. See here for supported test environments.
Example: if you run your test suite with parallelization and request 5 firefox, 5 chrome, 5 internet explorer on windows 10, then 15 windows 10 grid nodes are launched automatically if needed and registers with your grid hub to process your tests.
It takes on average about 2-7 minutes before an On-Demand grid node is launched and starts processing your tests, and about 15-30 seconds extra before a Spot grid node starts processing your tests. These launch times depend on how busy Amazon is at the time of the request and can vary at peak times. Use the pre-launch API to avoid/reduce delays before your tests have nodes to execute on.
The auto scaling functionality also facilitates the termination of grid nodes and you can specify the node termination schedule upon starting a grid. Example: select 1 hour, first termination attempt at 59 minutes, select 2 hours, first termination attempt at 1 hour 59 minutes, select 3 hours, first termination attempt at 2 hour 59 minutes and so on.
If the node is busy with processing tests the termination attempt is aborted and instead attempted again at the next node termination interval. A shorter schedule can be used to minimize node run time. A higher value is useful if you want to minimize test start times.