High availability and automatic failover

Automatic takeover without split-brain apply.

Durable leases, fencing epochs, capability-aware scheduling, and transferable checkpoints let a healthy agent resume a failed pipeline.

Plan my migration See HA & DR
Lease
Durable ownership

Record which agent currently owns the pipeline.

Fence
Epoch-guarded apply

Old owners cannot write after takeover.

Resume
Durable positions

Recover source and sink progress before running.

Control-plane HA

Schedule from durable shared state.

Ownership, desired state, agent capability, and failover history survive restarts. Work moves only through an expired lease or fenced switchover.

✓ Capability-aware agent selection
✓ Lease renewal and expiry telemetry
✓ Planned switchover and automatic takeover
✓ One merged status after reassignment

Data-plane safety

Resume after identity and timeline checks.

The new owner restores durable state, validates source identity, acquires the next epoch, and then resumes. Ambiguous timelines stop loudly.

✓ Remote checkpoint and spool recovery
✓ Source promotion continuity by engine
✓ Sink-side stale-epoch rejection
✓ Crash and partition recovery tests

Keep exploring

See the connected parts of the platform.

HA & DR Solution →

Translate mechanics into RPO and RTO.

Reliability →

Inspect local durability.

Control Plane →

Operate ownership and failover.

Keep one safe owner through every failure.

Tell us the agent fleet, database topology, and recovery objectives. We will map leases and failover.

Plan my migration See HA & DR