Browser tests are known to be the flakiest ones. This is partly because browser infrastructure is complicated to maintain. But the second reason is – mainstream browser automation tools such as Selenium server are far from being efficient.
A year ago I have shown Selenoid - a truly efficient replacement of the standard Selenium server. This year I would like to demonstrate how to organize a fault-tolerant and easily scalable Selenium cluster using virtual machines in the cloud. I will start by setting up several Selenoid nodes and configure them to send logs and recorded videos to S3-compatible storage. Then I will run multiple Ggr load balancer instances allowing to use all running Selenoid nodes and organize a single entry point to the cluster. Finally, we'll discuss how to work with VNC and video recording in such a cluster.