From deployment to production, why continuous delivery is the key
Have you ever found manual deployment or production of an environment a painstaking process that’s slow, error-prone and has no rollback? Do you think manual verification is troublesome? These are just a few of the problems in software development that Vikas Naiyar and Pranjal Kumar from Harness spoke about during a workshop at Future of Work 2019.
Vikas, Principal Engineer, Harness and Pranjal, Staff Software Engineer, Harness, led the attendees through the process of building a complex pipeline quickly, doing continuous verification of deployment using Machine Learning, managing and monitoring intelligence and so on. The workshop offered the attendees a chance to get hands-on experience on how to deploy web-based application on Kubernetes cluster through the demo.
Watch this Future of Work workshop on ‘Continuous delivery made simple’.
Vikas began by outlining the common problems that one faces in the deployment process including manual verification, consistent errors, and went on to explain how Harness solves these problems. “Harness has come up with a single stop shop that automates the whole continuous deployment pipeline and verifies deployment using Machine Learning,” he said.
The solution ensures that everything is automated, and deployments are faster. This helps save both time and money without the developer. When people create custom scripts for each and every step of deployment, it is difficult to maintain and there is no certainty about whether or not it will work. Also, it doesn’t give a consistent experience for the developers. Harness solves this problem by using multiple tools to build binary and multiple steps for deployment. They take over after an artifact has been built and take it directly to production, deploy it, verify and ensure that it is monitored every step of the way.
Some of the features of Harness cluster includes templatising the workflow so that they can use the same workspace in various places. By default, their application also has alert mechanisms where rules can be set. They support mail, slack and UI, and have on-premise installation as well.
Continuous verification using ML
Pranjal continued talking about how continuous verification works with Harness, where the company’s safety net reduces failure by 90 percent. “The bread and butter is the deployment, but the powerhouse is Harness where continuous verification is provided,” he said.
Soon after the deployment, they collect various metrics and categorise them based on performance, security, quality and so on. In manual cases, there could be an anomaly, which is an issue. But with Harness, the application automatically detects issues in production and lets you roll back. No custom logic or code is required, as the integration is automated. The advantage with Harness is that you don’t have to switch tabs as this is one single platform for deployment, verification, monitoring and management.
The company also ensures that the application is safeguarded 24x7. “Once the application is deployed, a metric is generated every minute, analysis is done using ML and the results are showcased,” said Pranjal.