Apps behind an nginx auth_request (Orchest / Intelligence / Process) rely on
the ddf-* cookies and on GET /auth/me. Several problems made them lose the
session or show 500 pages:
- Cookies were written with a hard-coded Domain=.dadosfera.ai in every
environment, so a staging login overwrote the production session (and
vice versa) in the same browser. Names now take AUTH_COOKIE_SUFFIX and the
domain comes from COOKIE_DOMAIN (staging: -stg / .stg.dadosfera.ai).
- maxAge was the token's absolute epoch (~56 years). It is now a real
duration, SESSION_COOKIE_MAX_AGE_DAYS (default 30).
- /auth/me answered 500 when a refresh failed for anything but
NotAuthorizedException, when the refreshed token failed validation, or
when DUC hung (no deadline). It now answers 401 and expires the dead
cookies for definitive failures, 503 for DUC transport errors/timeouts
(AUTH_RPC_TIMEOUT_MS, default 10s), keeps 403 answers instead of masking
them with a refresh, refreshes when only the refresh cookie is left, and
stops logging request headers (they contain the cookies).
- /auth/refresh-access-token rewrites all three cookies from the session the
frontend just proved valid, healing stale cookies from another login or
environment. /auth/sign-out expires all three and always responds.
Helm values wire the new variables for prd and stg.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AUEXmydTBX9G2CGJ97YQKL
- Add auth_provider to values.yaml (cognito/keycloak)
- Add AUTH_PROVIDER env var to deployment template
- Note: maestro only needs to know which provider is used, duc handles connection
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add auth_provider config to values.yaml (cognito/keycloak)
- Add AUTH_PROVIDER env var to deployment template
- Add conditional Keycloak env vars (URL, realm, client ID, secret)
- Enables parallel deployments with different auth providers
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add ELASTICSEARCH_URL and ELASTICSEARCH_API_KEY env vars
- Pull credentials from {env}/microservices/elasticsearch secret
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Change TLS check from ENV !== 'local' to REDIS_TLS === 'true'
- Add REDIS_TLS env var to Helm deployment template
- Default to true for OCI Redis, can override for self-hosted K8s
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Resolve ingress.yaml timeout annotation conflicts by keeping both configurations
- Resolve stg.yaml conflicts by merging affinity/resources config with restricted_ip
- Fix chart references in unimed-maestro and private-maestro sections
- Update all chart paths from ../maestro to ../helm-chart for consistency
This merge brings in the latest beta changes while preserving our Oracle migration
and affinity standardization configuration.
Oracle Migration Changes:
- Create deploy/helm-chart/ directory structure (moved from maestro/)
- Create deploy/helmfiles/ directory structure (moved from helmfiles/)
- Update chart references in helmfiles from ../maestro to ../helm-chart
- Update deploy-manually.yml from Azure AKS to Oracle OKE
- Update validate-k8s.yml from Azure AKS to Oracle OKE
- Replace Azure CLI with OCI CLI installation and configuration
- Replace Azure authentication with OCI authentication using secrets
- Replace az aks get-credentials with oci ce cluster create-kubeconfig
- Update helmfile paths from helmfiles/ to deploy/helmfiles/
- Remove DockerHub integration (push_to_dockerhub input and related steps)
- Change runner from [self-hosted, prd-azure] to [self-hosted, prd-oracle]
- Add environment field for proper Oracle deployment environment handling
- Add HOME environment variable for OCI CLI
Affinity/Resources Standardization:
- Add affinity configuration with name=general node selector in values.yaml
- Update deployment template to use conditional affinity and resources blocks
- Configure staging to disable both affinity and resources (null values)
- Enable production to use affinity targeting name=general nodes
- Enable production to use standard resource limits (100m-2000m CPU, 1500Mi-2Gi memory)
This brings maestro into full compliance with Oracle migration requirements and
implements the standardized affinity/resources pattern consistent with in-factory and duc.