Table of Contents

Selenium Grid
Selenium Grid is a hub-nodes architecture proxy server. The server will act as the Hub. It facilitates parallel execution of test scripts where one can configure different OS or browsers. Those are called nodes. Tests contact the hub to obtain access to remote browser instances. The Hub takes the requests from test scripts and then sends them to nodes based on the browser configuration.
The below image shows how selenium works:

Below image shows how the selenium grid works:

For deploying a selenium grid, Docker needs to be pre-installed on a server. There are multiple ways to deploy based on the requirement. For example, if only one server is used and smaller number of test cases need to run in parallel, the entire selenium grid can be deployed in one server.

If the requirement is to run multiple tests in parallel and there is a need for scalable selenium grid infrastructure, one can distribute docker containers across multiple servers. At the moment the docker swarm is being used by us to maintain the grid setup.

Step 2: Configure the infrastructure Introduction to docker swarm:A Docker Swarm is a group of either physical or virtual machines running the Docker application and configured to join in a cluster. Once a group of machines has been clustered together, the usual Docker commands can be executed, but the machines in the cluster will now carry them out. A swarm manager controls the cluster’s activities, and machines that have joined the cluster are referred to as nodes. One of the key benefits associated with the operation of a docker swarm is the high level of availability offered for applications. In a docker swarm, there are typically several worker nodes and at least one manager node responsible for handling the worker nodes’ resources efficiently and ensuring that the cluster operates efficiently. In automation, one can use the same machine as manager and node or have as many nodes and managers as per need. So for deploying selenium grid a docker-compose file needs to be built where all the configurations for docker images have to be specified. Below is the format of the sample docker file and an explanation of the parameters.

This is the format of the docker-compose file. Here there are multiple services for selenium hub and selenium nodes. In the above example, there is one selenium hub service and one selenium chrome node service. Image : Teams must specify the name of the Image the service needs. Here selenium/hub is the Image for deploying the selenium hub container. Ports : They can map ports from one container to another and containers to host as well. Volumes : They can map host volumes to container volumes to access files from the host inside the container or to access files from the container to the host. Environment : They can specify various environment parameters required for deploying a grid. Replicas : Under chrome service, they can mention how many replicas they need to deploy. They can create as many replicas as they want based on machine configuration. Placement : They can specify the placement of a particular service. For example, in the above docker-compose file, they are specifying placement for selenium hub in manager. If they do not specify it, the swarm will deploy hub in any swarm node or manager.

1.1 On Local Machine

The entire selenium grid infrastructure can be deployed in a single machine. This can be done by using the docker swarm and the above docker-compose file. Here a single machine will act as both manager and node. Below are the steps for the setup. 1 . First, one needs to initialize the docker swarm by using the below command. This command will initialize the docker swarm, and a selenium gid can be deployed into the swarm.

docker swarm init

Now the stack needs to be deployed by using the below command. In the below command, one needs to specify the path of compose file and the name they want to give to the stack.

docker stack deploy –compose-file=docker-compose.yml selenium

3 . Now they can verify if the stack is deployed or not by using the below command. This will display all deployed services. In this case, selenium hub and selenium chrome nodes. It will display the count of replicas of a particular service.

docker service ls

1.2 On Multiple Servers

If the need is to deploy selenium grid across multiple servers, multiple servers can be used as managers and nodes. Below are the steps to configure it in multiple servers.1 . First, teams need to go to the manager node they want to use as a manager and run the below command. When this command is executed, a command is generated that starts with Docker join as an output.

docker swarm init

2 . Copy docker joins command from the output of the above command and runs it in node servers. Now all the nodes will be part of the docker swarm. Status and nodes connected to the manager can be verified by using the command below in the manager node.

docker node ls

The stack can be deployed by using the below command. In the below command, the path of compose file and the name of the stack need to be specified.

docker stack deploy –compose-file=docker-compose.yml selenium

3 . This will distribute all the replicas across multiple nodes instead of having them in only one server. The size of the selenium grid can be increased by just modifying the replicas count in compose file. Step3 : After deploying the selenium grid, the grid can be verified by opening the below URL in the browser. http://:4444

Step 4 : Run the scripts. One can implement parallel execution at the framework level and run multiple tests in parallel by creating a Remote Web Driver pointing to the docker swarm manager. Step5 : To remove the entire selenium grid stack, one must run the below command.

docker stack rm selenium

Few advantages:

• Easily scalable automation infrastructure.
• Reduced cost for resources like servers, multiple VMs etc.
• Easy to deploy in multiple servers.
• There entire infra can be made open source and c Windows/Linux servers can be used.

Solution provided: Our team started analyzing the use of Docker, and as each container runs separately, they could deploy many docker containers at once in a machine. By using the docker swarm, they could deploy containers on multiple servers. They started with two servers and set up a docker swarm in those two servers. Now they could deploy as many containers as they wanted using docker swarm based on our server configurations. They deployed 10 selenium node containers per server and a hub in the manager server. In this way, they managed to run 20 tests in parallel. This worked fine, and execution time decreased drastically. Then they took another great step of horizontal scaling and increased the servers. If they increased one server at a time, theywould get 10 more parallel selenium nodes to run tests. Benefits: In this way, finally, they used four servers and managed to run 40 tests in parallel. Execution time decreased to below one hour for all the tests, and the client was able to run tests on multiple environments for each deployment to identify the failures faster.