Backstage on Docker with GitLab authentificaiton
Few days ago, I decided to finally get some time to discover #Backstage and understand the main concepts and how to use it.
In this article, I’ll explain the steps (and ambushes 😉) I followed to install it locally running on #Docker and with authentication based on #GitLab
Follow the official doc
So to begin, I’ve tried to follow the official documentation: https://backstage.io/docs/getting-started/
First, we need to create our app
$ npx @backstage/create-app@latest
? Enter a name for the app [required] hashnode-article
Creating the app...
Checking if the directory is available:
checking hashnode-article ✔
Creating a temporary app directory:
Preparing files:
copying .dockerignore ✔
templating .gitignore.hbs ✔
templating .eslintrc.js.hbs ✔
copying README.md ✔
...
executing yarn install ✔
executing yarn tsc ✔
🥇 Successfully created hashnode-article
All set! Now you might want to:
Run the app: cd hashnode-article && yarn dev
Set up the software catalog: https://backstage.io/docs/features/software-catalog/configuration
Add authentication: https://backstage.io/docs/auth/
Good, we are ready to go ! We can test that locally it works
$ yarn dev
yarn run v1.22.22
$ concurrently "yarn start" "yarn start-backend"
$ yarn workspace app start
$ yarn workspace backend start
$ backstage-cli package start
$ backstage-cli package start
[0] Loaded config from app-config.yaml
[0] <i> [webpack-dev-server] Project is running at:
[0] <i> [webpack-dev-server] Loopback: http://localhost:3000/, http://[::1]:3000/
[0] <i> [webpack-dev-server] Content not from webpack is served from '/private/tmp/hashnode-article/packages/app/public' directory
[0] <i> [webpack-dev-server] 404s will fallback to '/index.html'
[0] <i> [webpack-dev-middleware] wait until bundle finished: /
...
[0] <i> [webpack-dev-middleware] wait until bundle finished: /
[0] webpack compiled successfully
It works !
Ok, but this is not very convenient to share this with your team. Let’s ship it with Docker, it would be easier to deploy.
Package with Docker
First surprise : there is no official Docker image available (you’ll understand why later in this article 😉). So let’s follow follow the doc again to build our own image : https://backstage.io/docs/deployment/docker
First, we need to build the application
yarn build:backend --config ../../app-config.yaml
Now, we can build our image
docker image build . -f packages/backend/Dockerfile --tag backstage:demo
Let’s start it
$ docker run -it -p 7007:7007 backstage:demo
Loading config from MergedConfigSource{FileConfigSource{path="/app/app-config.yaml"}, FileConfigSource{path="/app/app-config.production.yaml"}, EnvConfigSource{count=0}}
{"level":"info","message":"Found 0 new secrets in config that will be redacted","service":"backstage"}
{"level":"info","message":"Listening on :7007","service":"rootHttpRouter"}
{"level":"info","message":"Plugin initialization started: 'app', 'proxy', 'scaffolder', 'techdocs', 'auth', 'catalog', 'permission', 'search', 'kubernetes'","service":"backstage","type":"initialization"}
/app/node_modules/@backstage/backend-defaults/dist/database.cjs.js:457
throw new Error(
^
Error: Failed to connect to the database to make sure that 'backstage_plugin_app' exists, Error: connect ECONNREFUSED 127.0.0.1:5432
at PgConnector.getClient (/app/node_modules/@backstage/backend-defaults/dist/database.cjs.js:457:15)
at runNextTicks (node:internal/process/task_queues:60:5)
at listOnTimeout (node:internal/timers:538:9)
at process.processTimers (node:internal/timers:512:7)
Node.js v18.20.4
Oops, even with the default config, we need a postgres database. No big deal, we can start one with Docker
$ docker network create backstage
$ docker run -d --name postgres -p 5432:5432 -e POSTGRES_HOST_AUTH_METHOD=trust --restart always --network backstage postgres
Note that we link it to a backstage
network in Docker
Let’s retry to start backstage
$ docker run --rm -it -p 7007:7007 -e POSTGRES_HOST=postgres -e POSTGRES_USER=postgres --network backstage backstage:demo
...
ForwardedError: Plugin 'auth' startup failed; caused by Error: Failed to initialize guest auth provider, The guest provider cannot be used outside of a development environment
at /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:677:19
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
... 3 lines matching cause stack trace ...
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5) {
cause: Error: Failed to initialize guest auth provider, The guest provider cannot be used outside of a development environment
at bindProviderRouters (/app/node_modules/@backstage/plugin-auth-backend/dist/index.cjs.js:1544:17)
at createRouter (/app/node_modules/@backstage/plugin-auth-backend/dist/index.cjs.js:2267:3)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async Object.init [as func] (/app/node_modules/@backstage/plugin-auth-backend/dist/index.cjs.js:2335:24)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
at async Promise.all (index 4)
at async #doStart (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:630:5)
at async BackendInitializer.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:556:5)
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5)
}
Hum, it seems that we have to setup a authentication provider, guest
mode is not authorized by default. As I have the chance to be a GitLab Hero, let’s try to connect with GitLab as there is a provider for it.
Authentication with GitLab
So, let’s follow the official documentation : https://backstage.io/docs/integrations/gitlab/locations
We have to update the app-config.yaml
file to add GitLab integration
...
integrations:
gitlab:
- host: gitlab.com
token: ${GITLAB_TOKEN}
Let’s rebuild our image to include this new configuration and try to re-run it (adding the GitLab token as environment variable)
$ docker image build . -f packages/backend/Dockerfile --tag backstage:demo
...
$ docker run --rm -it -p 7007:7007 -e POSTGRES_HOST=postgres -e POSTGRES_USER=postgres -e GITLAB_TOKEN=<your_token> --network backstage backstage:demo
...
MigrationLocked: Plugin 'catalog' startup failed; caused by MigrationLocked: Migration table is already locked
at /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:677:19
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
... 3 lines matching cause stack trace ...
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5) {
cause: MigrationLocked: Migration table is already locked
at /app/node_modules/knex/lib/migrations/migrate/Migrator.js:343:13
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async Migrator._runBatch (/app/node_modules/knex/lib/migrations/migrate/Migrator.js:359:7)
at async applyDatabaseMigrations (/app/node_modules/@backstage/plugin-catalog-backend/dist/cjs/CatalogBuilder-C7ANIkk3.cjs.js:1481:3)
at async CatalogBuilder.build (/app/node_modules/@backstage/plugin-catalog-backend/dist/cjs/CatalogBuilder-C7ANIkk3.cjs.js:6982:7)
at async Object.init [as func] (/app/node_modules/@backstage/plugin-catalog-backend/dist/alpha.cjs.js:242:46)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
at async Promise.all (index 5)
at async #doStart (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:630:5)
at async BackendInitializer.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:556:5)
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5)
}
Oops, looks like a bug 🙃 : https://github.com/backstage/backstage/issues/24284
So let’s do one of the workarounds : drop the database and re-run backstage
$ docker rm -f postgres
$ docker run -d --name postgres -p 5432:5432 -e POSTGRES_HOST_AUTH_METHOD=trust --restart always --network backstage postgres
$ docker run --rm -it -p 7007:7007 -e POSTGRES_HOST=postgres -e POSTGRES_USER=postgres -e GITLAB_TOKEN=<your_token> --network backstage backstage:demo
...
ForwardedError: Plugin 'auth' startup failed; caused by Error: Failed to initialize guest auth provider, The guest provider cannot be used outside of a development environment
at /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:677:19
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
... 3 lines matching cause stack trace ...
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5) {
cause: Error: Failed to initialize guest auth provider, The guest provider cannot be used outside of a development environment
at bindProviderRouters (/app/node_modules/@backstage/plugin-auth-backend/dist/index.cjs.js:1544:17)
at createRouter (/app/node_modules/@backstage/plugin-auth-backend/dist/index.cjs.js:2267:3)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async Object.init [as func] (/app/node_modules/@backstage/plugin-auth-backend/dist/index.cjs.js:2335:24)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
at async Promise.all (index 4)
at async #doStart (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:630:5)
at async BackendInitializer.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:556:5)
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5)
}
Then we need to specify the provider configuration : https://backstage.io/docs/auth/gitlab/provider/#configuration
😱 OMG: Need to modify the sources !! I understand know, why there is no official Docker image : when you want to enable a plugin, you’ll need to modify the source code of your application and re-build the Docker image …
Ok, let’s follow the doc and modify the source code as explain. Then modify again the app-config.yaml
to include provider information :
...
auth:
environment: local
providers:
gitlab:
local:
clientId: <your_gitlab_app_client_id>
clientSecret: <your_gitlab_app_client_secret>
audience: https://gitlab.com
signIn:
resolvers:
- resolver: emailMatchingUserEntityProfileEmail
- resolver: usernameMatchingUserEntityName
...
catalog:
providers:
gitlab:
yourProviderId:
host: gitlab.com
orgEnabled: true
group: fun_with # One of my groups on gitlab.com
allowInherited: true
groupPattern: '[\s\S]*'
schedule:
frequency: { minutes: 3 }
timeout: { minutes: 3 }
We re-build our image and run it (and probably the postgres container also…)
$ yarn build:backend --config ../../app-config.yaml
...
$ docker image build . -f packages/backend/Dockerfile --tag backstage:demo
...
$ docker run --rm -it -p 7007:7007 -e POSTGRES_HOST=postgres -e POSTGRES_USER=postgres -e GITLAB_TOKEN=<your_token> --network backstage backstage:demo
...
ForwardedError: Plugin 'kubernetes' startup failed; caused by Error: Kubernetes configuration is missing
at /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:677:19
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:11
at async Promise.all (index 8)
... 2 lines matching cause stack trace ...
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5) {
cause: Error: Kubernetes configuration is missing
at KubernetesBuilder.build (/app/node_modules/@backstage/plugin-kubernetes-backend/dist/index.cjs.js:1576:15)
at Object.init [as func] (/app/node_modules/@backstage/plugin-kubernetes-backend/dist/alpha.cjs.js:143:42)
at /app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:676:33
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async Promise.all (index 8)
at async #doStart (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:630:5)
at async BackendInitializer.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:556:5)
at async BackstageBackend.start (/app/node_modules/@backstage/backend-app-api/dist/index.cjs.js:832:5)
}
Oops, looks like another bug 🙃 : even if you don’t want to enable Kubernetes management in your instance, you have to provide configuration.
Found a workaround : https://github.com/backstage/backstage/pull/26223
Add this configuration to the app-config.yaml
:
...
kubernetes:
serviceLocatorMethod:
type: 'multiTenant'
clusterLocatorMethods:
- type: 'config'
clusters: []
Let’s go for another round of image building and run
$ docker image build . -f packages/backend/Dockerfile --tag backstage:demo
...
$ docker run --rm -it -p 7007:7007 -e POSTGRES_HOST=postgres -e POSTGRES_USER=postgres -e GITLAB_TOKEN=<your_token> --network backstage backstage:demo
...
{"level":"info","message":"Listening on :7007","service":"rootHttpRouter"}
...
🎉 it starts ! Let’s check it in a browser :
Sounds good but ...
Impossible to connect 🤨
Backstage based on authentication provider should create the users from the group I’ve defined in the configuration (the fun_with
one). But looking at the logs of the container, it seems that Backstage doesn’t find any user
Need to import users from a gitlab
https://backstage.io/docs/integrations/gitlab/org
Don't understand why, let's debug. Looking at the logs of the container, we can see that Backstage don’t find any users
{"class":"GitlabOrgDiscoveryEntityProvider","level":"info","message":"Scanned 0 users and processed 0 users","plugin":"catalog","service":"backstage","target":"GitlabOrgDiscoveryEntityProvider:yourProviderId","taskId":"GitlabOrgDiscoveryEntityProvider:yourProviderId:refresh","taskInstanceId":"da68ba04-34f3-4796-9050-526e598121b7"}
{"class":"GitlabOrgDiscoveryEntityProvider","level":"info","message":"Scanned 14 groups and processed 0 groups","plugin":"catalog","service":"backstage","target":"GitlabOrgDiscoveryEntityProvider:yourProviderId","taskId":"GitlabOrgDiscoveryEntityProvider:yourProviderId:refresh","taskInstanceId":"da68ba04-34f3-4796-9050-526e598121b7"}
Let’s deep dive into Backstage code and find that is not working for public groups on gitlab.com (NB : because seats concept is only for paid organization)
return this.listGroupMembers(groupPath, {
...options,
show_seat_info: true,
}
So I created an Issue on Github to ask for a fix. (After some discussions, first it was defined as the normal behavior, but after some exchanges, a fix will be applied to have the possibility to disable this flag)
So let’s try to switch to my own GitLab instance, I need to adapt the configuration
...
integrations:
gitlab:
- host: ${GITLAB_HOST}
token: ${GITLAB_TOKEN}
apiBaseUrl: https://${GITLAB_HOST}/api/v4
baseUrl: https://${GITLAB_HOST}
...
catalog:
providers:
gitlab:
yourProviderId:
host: ${GITLAB_HOST}
orgEnabled: true
groupPattern: '[\s\S]*'
schedule:
frequency: { minutes: 3 }
timeout: { minutes: 3 }
And let’s build and run a last (?) time our Docker image
$ yarn build:backend --config ../../app-config.yaml
...
$ docker image build . -f packages/backend/Dockerfile --tag backstage
...
$ docker run --rm -it -p 7007:7007 -e POSTGRES_HOST=postgres -e POSTGRES_USER=postgres -e GITLAB_TOKEN=<your_token> --network backstage backstage:demo
...
{"class":"GitlabOrgDiscoveryEntityProvider","level":"info","message":"Scanned 5 users and processed 5 users","plugin":"catalog","service":"backstage","target":"GitlabOrgDiscoveryEntityProvider:yourProviderId","taskId":"GitlabOrgDiscoveryEntityProvider:yourProviderId:refresh","taskInstanceId":"3792184e-6eac-412c-968c-9cf02041b6f1"}
...
The logs looks better, let’s try to connect
Finally !
In summary…
After some little troubles, I finally succeed to install Backstage on a Docker image with a GitLab provider for authentication. Now my instance is fully operational.
What have I learn during my first experience with Backstage is that :
it seems very customizable
but
you have to modify the sources to enable plugins and some features
there are some bugs and little issues
the team is super reactive, I had a feedback to my new issue less than 2 days after
So I’ll probably continue to explore Backstage because I want to understand how helpful it can be in companies to enable a developer portal but the fact that it’s not easily configurable with some configuration for plugins and to have to modify the source code each time is quite disturbing for me 😉
Subscribe to my newsletter
Read articles from Matthieu Vincent directly inside your inbox. Subscribe to the newsletter, and don't miss out.
Written by
Matthieu Vincent
Matthieu Vincent
TechAdvocate DevSecOps / Cloud platform Lead Architect @ Sopra Steria Speaker @ Devoxx, Snowcamp, Breizhcamp, GitLab Connect and internally Co-Founder of Volcamp IT Conference @ Clermont-Fd (https://volcamp.io) GitLab Hero