The DevOps toolchain is fragmented and creates "Islands of Automation" where specific activities are automated but the dependencies and interactions between them are not. Shippable solves this problem by providing an easy way to connect your islands of automation into continuous, streamlined, stateful workflows.
State(ful) workflows are designed to remember information generated by every job and resource and make it available to downstream jobs or successive executions of the same job. This information could be as simple as the status of the activity all the way up to data about access keys, IP addresses, network configurations, and so on. The information needs to be accessible from a systems perspective if you want to achieve Continuous Delivery.
Let's look at an example, a Continuous Integration activity would need to generate data about the commit SHA, Build Number, image tag, storage location of build output, test results, etc. This is necessary for the Test team to be able to deploy the right version of the service into test environments. If the Test team has to ask the Dev team for this information, it is not a frictionless process and will be error-prone.
Typical use cases for State are the following:
- Storing network configuration like VPC info, subnets, security groups, etc.
- List of security approved machine images for VM deployments
- Environment information like IP Addresses, Access Keys, Bastion hosts, etc.
- Storing output information from popular DevOps tools, e.g., Terraform output that needs to be present to rerun the activity
- Status of CI jobs like build number, commit SHA, tags, coverage reports, and test reports
- Release version list, time of release, components that make up the release with immutable tags
- Deployment state, time, versions, and so on
- Environment variables and configurations required to run the app
There are many more use cases where this applies, like service discovery, key-value stores, load balancers, and so on.
The Assembly Lines platform supports three forms of state:
Key-value state: is stored as a property in the immutable version of a resource or job. Every job that uses the resource will have access to the version and in turn has access to the key-value pair that was stored in that version.
File based state: is available only in jobs and state files up to 1MB can be stored for every version, i.e. run, of a job. Every subsequent job that has this as an input has access to those files. This makes it easy to transfer stateful data from one job to another.
Shared state resource: state always flows from left to right. However, there are some scenarios where a shared state is needed. For example, Terraform creates a statefile when
terraform applyis run. But if you need to run this in a multi-stage manner where subsequent jobs make changes to this state, it somehow needs to flow back to the first job that created it. Our Assembly Lines has a special shared
stateresource that allows this state to be stored centrally and shared across the jobs.