Platform 2026.05
Released May 2026
This is the initial platform release under the new YYYY.MM versioning scheme.
Components
| Component | App Version | Chart Version | Notes |
|---|---|---|---|
| Kubernetes | 1.33 | — | GKE, EKS, or AKS depending on your cloud |
| Airflow | 3.2.2 | 1.21.0 | Apache Airflow with Python 3.12 |
| Airbyte | 2.0.1 | 2.0.19 | Data integration / ELT platform |
| Superset | 6.1.0 | 0.15.5 | Data visualization and dashboards |
| DataHub | 1.5.0.6 | 0.11.0 | Data catalog and lineage |
| KEDA | 2.19.0 | 2.19.0 | Event-driven pod autoscaling |
| CloudNativePG | 1.29 | 0.28.2 | Managed PostgreSQL operator |
| Terraform | 1.15.3 | — | Infrastructure provisioning |
Highlights
Multi-cloud support
SkaleData now deploys to AWS (EKS) and Azure (AKS) in addition to GCP (GKE). All three clouds share the same application layer — Airflow, Airbyte, Superset, and DataHub work identically regardless of where your cluster runs.
- AWS — EKS with managed node groups, S3 storage, ECR container registry, Secrets Manager
- Azure — AKS with VMSS node pools, Blob Storage, ACR container registry, Key Vault
- GCP — GKE with node pools, GCS storage, Artifact Registry, Secret Manager
Custom Airflow images
Deploy your own Airflow image with custom Python packages and DAGs:
skale airflow deploy --tag my-image:v1.2.3The platform uses a stable :current tag with imagePullPolicy: Always — push a new image and restart to deploy. No Helm upgrade required for subsequent deploys.
DAG-only deploys
Push DAGs to your cluster without rebuilding the entire image:
skale airflow deploy --dag-onlyDAGs are synced via a sidecar container every 30 seconds from your cloud storage bucket.
Platform version abstraction
Cluster versions are now expressed as YYYY.MM (e.g., 2026.05) instead of internal Scaffold tags. The console, CLI, and API all display the platform version.
Per-instance secrets backend
Each Airflow instance gets its own isolated secrets backend:
- GCP — Secret Manager with prefix
airflow/{instance}/ - AWS — Secrets Manager with prefix
airflow/{instance}/ - Azure — Key Vault with dedicated vault per instance
Programmatic API access
Access your data plane applications (Airflow, Airbyte, Superset) programmatically via API keys:
curl -H "Authorization: Bearer sk_live_..." \
https://{app}.skaledata.run/api/v2/dagsHelm-native app lifecycle
Applications are managed directly via Helm (not Terraform). This means:
- Faster app provisioning (~2 min vs ~10 min)
- Independent app lifecycle (add/remove/restart without touching infra)
- Multi-instance support (run multiple Airflow instances per cluster)
Upgrade notes
This is the initial versioned release. All existing clusters are automatically on 2026.05.
Infrastructure changes
- Cloud-agnostic module architecture (11 modules renamed, shared orchestration extracted)
- NGINX ingress uses
externalTrafficPolicy=Localon Azure for proper health probes - EKS cluster autoscaler included by default
- gp3 StorageClass for EKS (replaces gp2)
- Default node sizes: GCP
e2-standard-4, AWSm5.xlarge, AzureStandard_D4s_v3