Multi-DC HA Kubernetes 🤖 MCP Native

Manganese ProdFleet HA

Production-grade high availability:
Multi-DC K8s · PostgreSQL streaming · DNS failover — without hyperscaler dependency

Technical foundations

Engineered for critical production -- not adapted from a dev platform

🌐

Kubernetes orchestration (Pack8s)

Multi-cluster Helm releases synchronised via sync_follower_with_primary. A real helm upgrade executes on followers -- not passive observation.

🗄️

Cross-DC PostgreSQL streaming

Active streaming replication over any IP network. pglookout monitors health. On failure, the standby takes over without backup restore.

🧠

Automatic health state machine

meta_cluster continuously calculates cluster state: healthy / degraded / critical / broken. Encoded actions execute without waiting for an operator.

🔀

Wired DNS failover

failover_mode flag on each DNS record. Active-passive switchover via btn_switchover. Automatic Sunray rule propagation without operator action.

📦

Multi-provider S3 storage

8 sovereign backends: MinIO, OVH, Scaleway, Cloudflare R2, AWS, Infomaniak, Wasabi, Dell ECS. Each bucket carries its own credentials.

🤖

MCP-native control

~35 typed MCP methods + dedicated HA operations. Your AI pipelines run in HA prod without rewriting from Claude, Cursor or any MCP client.

Two HA levels

From single-DC to multi-site active-passive

HA Single-DC

Pack K8s + pglookout

Health-driven, not manual. Add/remove instances via wizard. State calculated automatically.

HA Multi-DC

Pack8s + Helm cross-cluster

sync_follower_with_primary executes real helm upgrades. Wired DNS failover. Operational active-passive switchover.

Why ProdFleet HA

Real HA, not a toggle switch

The state machine continuously calculates cluster state and triggers encoded actions -- without waiting for an operator.

Replicated PostgreSQL, zero cross-DC loss

Active streaming replication over standard IP network, monitored by pglookout. On datacentre failure, the standby takes over without backup restore.

Kubernetes without proprietary lock-in

Standard Helm, cross-cluster sync via sync_follower_with_primary: your workload stays portable to any third-party K8s cluster.

MCP operational continuity

Same agent control surface as DevFleet: your AI pipelines run in HA prod without rewriting, with added HA methods.

Total sovereignty, including on-prem

Muppy Enterprise control plane installable on your hardware via Docker. Zero Odoo licence, zero mandatory US cloud, 8 S3 providers available.

Single-author cohesion

Manganese, Sunray and IMQ are developed by the same team. Cross-component edge cases are resolved internally, without fragile third-party integrations.

Pricing

Beta phase -- pricing coming soon on all plans

Managed by oursbl.eu

Dedicated Public

Beta Version

Pricing coming soon

Dedicated server leased and managed by oursbl.eu *
Orchestrated Kubernetes (Pack8s)
PostgreSQL HA included
S3 multi-provider included
Sunray Zero Trust included
Full MCP access

* Provider rate +20%

Choose Dedicated

Your hardware

Dedicated Owned Public

Beta Version

Pricing coming soon

You bring the public server or VPS
Oursbl.eu enrols and manages the K8s platform
PostgreSQL HA included
Sunray Zero Trust included
Full MCP access
Zero markup on hardware

On-Premise

Muppy Enterprise

Beta Version

Pricing coming soon

Control plane installed at client site
Portable Docker install (Muppy Enterprise)
Managed K8s data plane included
Zero Odoo licence required
Target: CIO / Mid-market / Public sector
0.20 EUR/GB RAM beyond floor

* Setup billable per day

Air-gap . Beta Q3 2026

On-Premise Managed

TBD

beta Q3 2026

Air-gapped infrastructure at client site
Managed from oursbl.eu SaaS via mpy-tunnel
Kubernetes on your private datacentre
Zero mandatory outbound flow
Target: public sector / OIV

Full HLD, beta Q3 2026

Join the beta

Ready for genuine production HA?

Multi-DC · K8s · PostgreSQL streaming · Zero hyperscaler dependency