Added
- (flux-react) Show each resource card’s manifest as YAML by @marians — In the Flux UI’s details panel, every resource card has a
View YAML button that opens the resource’s manifest read-only, with Copy manifest. The agent and workflow manifest dialogs look and copy the same way. - (gs) Show worker CPU and memory capacity of clusters by @QuentinBisson — A cluster’s About card shows
Worker capacity, e.g. 4 nodes · 32 vCPUs · 128 GiB RAM, and the Clusters table has optional, sortable Worker CPU and Worker memory columns. - (plans) Hive merges Plans, Roadmap and the magazine by @teemow — Plans, Roadmap and the product magazine become one sidebar entry, Hive, with the tabs Now, History, Roadmap, Plans and Knowledge and a shared team selector and search. Old links redirect. Disabled by default.
Changed
Fixed
- (agent-platform) Read delete_cluster’s nothingLeft as removal done by @teemow — The
Delete cluster dialog ends in a success message once nothing of the cluster is left, instead of a red refusal, also when opened on a cluster that is already removed. - (agent-platform) Show Delete only on clusters create_cluster made by @marians — On a cluster’s page,
Delete shows only for clusters made with Create cluster. Other clusters no longer offer it, where the dialog could only show a refusal. - (app) Skip TelemetryDeck on an empty app ID, log init failures by @teemow — For operators: with an empty TelemetryDeck app ID, the portal logs one browser console warning and sends nothing, and a failed initialisation is logged instead of silently dropping all usage data.
- (app) Offer the GitHub tabs only on a portal with a GitHub login by @teemow — On portals without a GitHub login, a component’s page no longer shows the GitHub Actions and Pull Requests tabs, which only opened a
Login Required: GitHub dialog that could never succeed. - (backend) Warn at startup when the telemetry app ID is empty by @teemow — For operators: with TelemetryDeck enabled but
app.telemetrydeck.appID empty, the backend logs a warning at startup, since the portal then sends no usage data at all. - (backend) Retry the database at start, exit on a failed boot by @teemow — For operators: a backend started before its database retries for up to five minutes and turns ready once the database is up. A failed start exits instead of hanging unready, and the chart adds a
startupProbe. - (gs) Keep the sidebar without a main auth provider by @teemow — On a portal without a main auth provider, such as a local start with guest sign-in, the sidebar renders again instead of disappearing.
- (gs) Open the cluster page of a cluster a HelmRelease installs by @teemow — The page of a cluster installed by a Flux HelmRelease, such as one made with
Create cluster, opens instead of showing a NotFoundError, with a new Installed by card showing the release, chart, version and status. - (gs, kubernetes-react) Secret store picker lists external-secrets.io/v1 stores and shows readable errors by @marians — In templates such as RDS and ElastiCache, the secret store picker lists SecretStores and ClusterSecretStores again, and says why when it can’t, instead of
[object Object]. - (scaffolder) Refresh cluster tokens when a template is submitted by @QuentinBisson — Templates that apply resources to a cluster no longer fail with an expired token after a long form: the token is fetched on
Create. A failed start shows an error on the review step, and Create works again.
Development and CI/CD
Full Changelog: https://github.com/giantswarm/backstage/compare/v2.91.1...v2.92.0