Registration now goes through POST /api/auth/register, which admin-creates the Kratos identity (unverified email, NO auto-sent code). Kratos self-service registration never returns the verification flow id, so the client could never submit the user's code to the right flow; admin creation lets the client own a single verification flow instead. Also surface the live Kratos verified flag and fix Apple audience + team IDs. - kratos.Client.CreateIdentity via admin API; ErrIdentityExists / ErrInvalidCredentials - AuthService.Register + AuthHandler.Register + public POST /api/auth/register/ - CurrentUser overrides stale user_profile.verified with the live Kratos flag; UserRepository.MarkVerified mirrors it back - configmap: additional_id_token_audiences allows the .dev bundle id_token - fix Apple/APNs team id V3PF3M6B6U -> X86BR9WTLD in .env.example + dev init Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
honeyDue — K3s Dev Deployment
Single-node K3s dev environment that replicates the production setup with all services running locally.
Architecture: 1-node K3s, in-cluster PostgreSQL + Redis + MinIO (S3-compatible), Let's Encrypt TLS.
Domains: devapi.myhoneydue.com, devadmin.myhoneydue.com
Quick Start
cd honeyDueAPI-go/deploy-k3s-dev
# 1. Fill in config
cp config.yaml.example config.yaml
# Edit config.yaml — fill in ALL empty values
# 2. Create secret files (see secrets/README.md)
echo "your-postgres-password" > secrets/postgres_password.txt
openssl rand -base64 48 > secrets/secret_key.txt
echo "your-smtp-password" > secrets/email_host_password.txt
echo "your-fcm-key" > secrets/fcm_server_key.txt
openssl rand -base64 24 > secrets/minio_root_password.txt
cp /path/to/AuthKey.p8 secrets/apns_auth_key.p8
# 3. Install K3s → Create secrets → Deploy
./scripts/01-setup-k3s.sh
./scripts/02-setup-secrets.sh
./scripts/03-deploy.sh
# 4. Point DNS at the server IP, then verify
./scripts/04-verify.sh
curl https://devapi.myhoneydue.com/api/health/
Prod vs Dev
| Component | Prod (deploy-k3s/) |
Dev (deploy-k3s-dev/) |
|---|---|---|
| Nodes | 3x CX33 (HA etcd) | 1 node (any VPS) |
| PostgreSQL | Neon (managed) | In-cluster container |
| File storage | Backblaze B2 | MinIO (S3-compatible) |
| Redis | In-cluster | In-cluster (identical) |
| TLS | Cloudflare origin cert | Let's Encrypt (or Cloudflare) |
| Replicas | api=3, worker=2 | All 1 |
| HPA/PDB | Enabled | Not deployed |
| Network policies | Same | Same + postgres/minio rules |
| Security contexts | Same | Same (except postgres) |
| Deploy workflow | Same scripts | Same scripts |
| Docker images | Same | Same |
TLS Modes
Let's Encrypt (default): Traefik auto-provisions certs. Set tls.letsencrypt_email in config.yaml.
Cloudflare: Same as prod. Set tls.mode: cloudflare, add origin cert files to secrets/.
Storage Note
MinIO provides the same S3-compatible API as Backblaze B2. The Go API uses the same env vars (B2_KEY_ID, B2_APP_KEY, B2_BUCKET_NAME, B2_ENDPOINT) — it connects to MinIO instead of B2 without code changes.
An additional env var STORAGE_USE_SSL=false is set since MinIO runs in-cluster over HTTP. If the Go storage service hardcodes HTTPS, it may need a small change to respect this flag.
Monitoring
stern -n honeydue . # All logs
kubectl logs -n honeydue deploy/api -f # API logs
kubectl top pods -n honeydue # Resource usage
Rollback
./scripts/rollback.sh