Changes and Releases
Updates on Giant Swarm workload cluster releases, apps, UI improvements and documentation changes.
Added
- (agent-platform) Change Harness on the agent page, and earlier-revision sessions marked by @QuentinBisson —
Change Harness… in an agent’s actions menu moves it to another runtime after showing whether it can run there. On the agent’s Sessions tab, sessions started before the change are marked Earlier revision. Needs agent-manager 2.7.0 or later. - (agent-platform, gs-backend) Plugins on the agent wizard’s skills step by @teemow — Where the skill repositories hold plugins, the new agent wizard’s step becomes
Select skills and plugins: a Claude Code plugin is picked whole, an Agent Plugins package skill by skill. Edit keeps an agent’s plugins. Needs agent-manager 2.6 or later.
Development and CI/CD
Full Changelog: https://github.com/giantswarm/backstage/compare/v3.12.0...v3.13.0
Changed
- (tests) No PolicyException matches on user info while background scans apply it in #1168 by @teemow
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v4.157.0...v4.157.1
Added
- (components) Select Substrate 1.17 (the actor JWT mint as MintActorJwt, agentgateway 2.6.5) in #1164 by @QuentinBisson
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v4.156.1...v4.157.0
Added
route.backendTrafficPolicy.timeout sets the Envoy Gateway timeouts of the Zot route. Without it Envoy’s
default 15s route request timeout applies to the whole response, so large blob pulls on management clusters
with a slow uplink fail with a 504. http.requestTimeout: 0s disables it.
Changed
route.backendTrafficPolicy.connection is part of values.schema.json, like the new timeout. Neither was
validated before, so a misspelled field such as bufferLimt rendered into the policy and was silently
dropped. Both now accept only the fields of Envoy Gateway v1.9: values with any other connection or
timeout key, which rendered before, now fail schema validation.
Fixed
podDisruptionBudget.minAvailable and podDisruptionBudget.maxUnavailable are usable again. The generated
values.schema.json typed both as null only, because values.yaml defaults them to null, so
helm install --set podDisruptionBudget.minAvailable=1 failed schema validation and no PodDisruptionBudget
could ever be created. Both now accept an integer, a percentage string or null.- The chart unit tests run and pass. They did not: all 65 errored, first on the schema defect above and then
because the deployment and statefulset suites restrict rendering to one template while both templates
include configmap.yaml and secret.yaml for their checksum annotations. - The
helm.sh/chart label is valid for long chart versions: the 63-character cut trims the whole trailing run of -, . and _.
Fixed
- (substrate) The Organizations’ workers get a PolicyException of their own, with background false in #1161 by @fiunchinho
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v4.156.0...v4.156.1
Added
- (components) Select kagent 1.21 (tls:// Agent egress, no loopback token) in #1160 by @teemow
Changed
Full Changelog: https://github.com/giantswarm/agent-platform/compare/v4.155.1...v4.156.0
Fixed
- Values schema: accept arbitrary keys again in free-form maps and pass-through Kubernetes objects (annotations, labels,
nodeSelector, affinity, security contexts, probes, resources, extraArgs as a map, tlsConfig, secretConfiguration.data), and name in imagePullSecrets items. This broke installs, for example with the IRSA eks.amazonaws.com/role-arn service account annotation on CAPA. - Values schema: accept the
aws values that cluster-aws still sets. They have no effect.
Changed
- Upgrade external-dns to v0.23.0.
- Sync to upstream helm chart 1.23.0.
Changed
- The release workflows authenticate as the
giantswarm-release GitHub App instead of the taylorbot user, regenerated with devctl v8.148.1 (giantswarm/giantswarm#2280).
