Steps Towards Automated Application Performance “Management” with Instana and Gimbal at Scale

Using Automatic Service Performance Monitoring to Drive Decision Making for Canary Deployment and A/B Testing in CI/CD

Heptio Gimbal enables efficient routing of traffic to applications deployed to multiple clusters such as Kubernetes and OpenStack. It is a perfect foundation for a mix of modern and classic stacks, as well as for stepwise transformation and migration of workloads from the classic world onto Kubernetes.

Especially in the case of migration it is very important to carefully do A/B tests and slow canary rollouts in order to ensure that the new application versions are performing at least as good as previous versions. Even more – it is very important to react on performance degradation and efficiently roll back to the previous version and address performance issues as soon as possible.

In order to achieve that, a high-resolution, continuous monitor needs to be observing service and application performance behaviour across mixed infrastructure. This creates a foundation data set upon which to compare the previous versions to newer applications through the rapidly growing CI/CD application development cycle. It’s also crucial that the monitor automatically discovers new workloads, attaches to them when needed and collects comprehensive performance data. This is particularly important at large scale. Gimbal itself has integrated endpoints that allow to monitor with Prometheus and visualise with Grafana, but at large scale this will require a more automated application monitoring capability.

That’s where Instana shines. It fully integrates into the enterprise CI/CD pipeline, automatically collects metrics and traces throughout the whole modern and classic stacks underneath and behind Gimbal and derives clear health statements per service or per whole applications – without configuration and rules maintenance. This allows DevOps teams to quickly react upon unpleasant changes and performance degradations and keep their users happy without service interruptions.

But with Gimbal, Instana doesn’t stop just there – together with the Gimbal community, Instana is architecting and building the feedback loop that allows the trigger of actions – be it rerouting, rollback or redeployment – based on the health information about the underlying services continuously computed by Instana. The foundation for that in Gimbal will be fully open and allows any kind of machine intelligence or human decision making to trigger these actions. But Instana is the first one of that kind to provide end-to-end fully automated feedback loop into Gimbal.

This combination of continuous, automated monitoring integrated with and informing dynamic routing into inhomogenous workloads clearly is the future of modern, large-scale enterprise operations. Maximum agility with non-stop feedback and remediation is where the world is heading, and Instana provides that future with Gimbal today.