Make Your Backstage Templates Resilient
Learn what actions you can follow from Backstage release v1.23.0-next.0 to make your templates resilient to redeployment or any accidental server failures.
Join the DZone community and get the full member experience.
Join For FreeI suppose that you are a user of Backstage and have your own templates that automate some repetitive and/or complicated flows of your organization. Some of your templates take a long time to be executed. For example, you provision your infrastructure, and your template guards the process along the way. Now, you are happy that you provided your engineers with these templates to lessen their burden, but you have to be careful with redeploying Backstage because you might fail ongoing tasks and make your engineers angry.
Here, I’ll explain what actions you can follow from release v1.23.0-next.0 to make your templates resilient to redeployment or any accidental server failures.
Steps
Since release v1.23.0-next.0 (or you can wait for a monthly release 1.23.0, which will happen at the end of February), you can mark your templates as recoverable, but you have to be sure that all your actions are idempotent. That means that your action can be running multiple times and having the same result as being called once.
Let’s go step by step what you have to do:
1. Enable recoverability in your app-config-local.yaml file:
scaffolder:
EXPERIMENTAL_recoverTasks: true
2. Open your template which you would like to make recoverable and there, add EXPERIMENTAL_recovery
config as follows:
apiVersion: scaffolder.backstage.io/v1beta3
kind: Template
metadata:
name: docs-mocks-idempotent-template
title: Documentation Mock Idempotent running Template
description: Create a new standalone documentation project
tags:
- recommended
- techdocs
- mkdocs
spec:
owner: backstage/techdocs-core
type: documentation
lifecycle: experimental
EXPERIMENTAL_recovery:
EXPERIMENTAL_strategy: 'startOver'
At the moment, only the startOver
strategy is supported. To switch it off, you can remove the EXPERIMENTAL_recovery
block or provide the value 'none'
.
3. Make each of your actions idempotent. Let's assume your template is simplistic and has the following actions inside:
steps:
- id: fetch
name: Template Docs Skeleton
action: fetch:template
input:
url: ./skeleton
values:
name: ${{ parameters.name }}
description: ${{ parameters.description }}
destination: ${{ parameters.repoUrl | parseRepoUrl }}
owner: ${{ parameters.owner }}
- id: publish
name: Publish
action: publish:github
input:
allowedHosts: [ "github.com" ]
description: This is ${{ parameters.name }}
repoUrl: ${{ parameters.repoUrl }}
- id: register
name: Register
action: catalog:register
input:
repoContentsUrl: ${{ steps.publish.output.repoContentsUrl }}
catalogInfoPath: "/catalog-info.yaml"
fetch:template
- This will be idempotent for your case, as at this moment, Backstage doesn't preserve the workspace; on each run, it will create a new one.publish:github
- This has to be replaced with your custom action. The modification to this action can look like this:
catalog:register
- This has to be also replaced with your custom action, and inside, prior to registering a new location, check if the location hasn't been registered yet. If yes, skip it, instead of throwing an error.
Give it a try. If you have more complicated templates, most probably you have more of your custom actions than default ones, and you are quite comfortable working with such customizations.
In future releases, you can find the most used default actions supporting idempotency, but creating more complicated templates will require homework from your side in any case.
Don't waste your time: start your journey now! Drop your feedback on what you think about it or if you need some more clarification on certain things as to how to make it fly on your end.
Opinions expressed by DZone contributors are their own.
Comments