Building the future of DevOps automation
Today, we’re starting Relay by Puppet, a new initiative to build the future of cloud-based DevOps automation. This post captures our musings about the evolving world of cloud-native infrastructure and how Relay addresses this evolution.
Working with many of the largest companies in the world, we’ve seen server virtualization increase the number of things that operators have to manage from 1x to 10x. The automation we’ve built today enables us to scale to large infrastructure estates and eliminate this manual work. We’ve made one-click automation possible.
However, in moving to public cloud, the number of things that operators have to manage has gone from 10x to 100x. A single VM running in a public cloud comprises a set of cloud resources that include the VM itself, the network and subnet it resides in, the IP address, the storage account, the disks, the security rules, the roles and permissions, and more. When you start to layer on container orchestration, the number of things to manage increases dramatically.
Similarly, the number of tools that operators use every day and tie together to solve automation challenges has exploded in number and complexity. Operators must learn infrastructure provisioning with Terraform, configuration management with Bolt, Kubernetes deployments with Helm, building pipelines with Jenkins, maintain a whole flock of public cloud services — all in addition to responding to the endless queue of ServiceNow tickets, demanding that someone restart their server for them.
In this world, human-triggered operations can no longer keep up. When you’re dealing with hundreds of tasks, even one-click automation is one click too many. Operators today need a different category of automation that responds to monitoring alerts, new service request tickets, cloud events and more. This kind of automation eliminates the low-value tasks operators spend their time doing today so that they can focus on the high-value work for their teams.
When you’re dealing with hundreds of tasks, even one-click automation is one click too many.
Relay is an event-driven automation service for eliminating low-value DevOps tasks. Relay makes it easy to pull together the tools you use today into a single workflow and automate these workflows with an event-based trigger.
We think there are a few useful scenarios here that either started out as problems for us or the DevOps engineers we talked to about Relay:
- Incident triggers: Enabling PagerDuty or VictorOps incident to trigger creating a war room in Slack
- Ticket triggers: Using a Jira ticket to trigger infrastructure provisioning of development environments with Terraform
- Git triggers: Deploying new versions of a containerized application to Kubernetes when a Github PR is merged
- Monitoring triggers: Clearing out /var/log when SignalFx throws a disk utilization alert
- Cloud event triggers: Making sure new S3 buckets are configured with the appropriate permissions when created
We’re not trying to boil the ocean here. Instead, we’re building a small set of use cases for DevOps teams that will make life way better for us (and hopefully you too).
We’re hard at work on our public beta launch (coming soon!) and we’d love to hear from you. What are the things that you’d like to automate today? Let us know. What’s a terrible idea (hello, storage policies should be separate from alerting)? Tell us in exchange for free swag!
Over the next couple months, we’ll talk through our thinking, our bets, our progress, and what we’re hearing from folks like you.
Let us know what to automate. We’ll give you free stuff!
Interested in following along? Join our mailing list and we’ll certainly offer our earliest supporters perks along the way (Kickstarter-style vague promises — check!).
Good ideas (and products) don’t happen in a vacuum. And really, no ideas are new. We’re inspired by the work that teams at Zapier, Stackstorm, Github Actions, and others have created before us. In the next few months, we’ll talk about how we see Relay being similar or different from many of these tools.