Execution Queue
Durable (DB-backed)
Capacity: 30,000 items
Locations and IPs
Queue-aware node routing across global and regional lanes
Uptimer schedules work with tenant fairness, bounded concurrency, and queue backpressure. Routing chooses the least-stress location when telemetry is fresh, then falls back predictably.
Node System at a Glance
Global Concurrency
48
Per-tenant cap: 6
Queue Snapshot
0 pending
Processing: 0, dead-letter: 0
Location Lanes
5
Node lanes: 2, global lanes: 3
Fairness and Concurrency Model
Tenant fairness in scheduler selection
Due monitors are bucketed per tenant and selected in round-robin order. This keeps one heavy tenant from consuming every scheduler batch.
Hard caps for stability
Worker execution is bounded by a global semaphore and a tenant semaphore. Backpressure is intentional: overload turns into queue depth, not process collapse.
Least-stress routing with fallback
Automatic routing uses fresh stress heartbeats to pick the least-stress location. If telemetry is stale or missing, routing falls back to deterministic lane rotation.
Reliable retry lifecycle
Durable mode supports lease recovery, exponential backoff retries, and dead-letter handling after max delivery attempts.
Locations, Lanes, and IP Planning
| Location Code | Lane Type | Routing Behavior | Node Stress |
|---|---|---|---|
global-1 |
Global Cloud Lane | Local cloud execution lane with stress-aware auto-selection. |
0.0% (fresh) |
global-2 |
Global Cloud Lane | Local cloud execution lane with stress-aware auto-selection. |
0.0% (fresh) |
global-3 |
Global Cloud Lane | Local cloud execution lane with stress-aware auto-selection. |
0.0% (fresh) |
node-1 |
Regional Node Lane | Remote node queue dispatch with local fallback on dispatch failure. |
0.0% (fresh) |
node-2 |
Regional Node Lane | Remote node queue dispatch with local fallback on dispatch failure. |
0.0% (fresh) |