Operations
Infrastructure health, queue pressure, capacity, and cost at a glance.
System Health
Image job flow, latency, queue pressure, and server freshness.
Overall
Image p90 E2E
Oldest Queued
Oldest Active
Ready Capacity
Failure Rate
Queue Freshness
Runway
Needs Attention
Image Jobs
| Window | Done | Failed | p90 E2E | p90 Queue | p90 Assign→Exec | p90 Ready→Exec | p90 Exec | p90 Upload | Max E2E |
|---|
Active Server Instances
Track live servers, mining, queue capacity, and instance controls.
Total Instances
Hourly Cost
PRL Earned 24h
NOCK Earned 24h
Vast Cost 24h
Mining Net 24h
Account Balance
Runway
Active Jobs
Utilization
Queued Jobs
PRL Mining
Dedicated rentals and idle mining on job workers.
Dedicated mining rentals
| Miner | Status | Hashrate | PRL/day | Revenue/day | Rent/day | Net/day | Margin | Running | Since rental |
|---|
Idle mining on job workers
| Worker | Status | Time mining (2h) | Hashrate | PRL/day | Revenue/day | Rent share/day | Net/day | Updated |
|---|
How these numbers are calculated
- PRL/day is hashrate × network yield (PRL per TH/s per day, from HeroMiners). It uses the pool's 1-hour worker rate when available, otherwise the miner's own reported hashrate.
- Dedicated rentals exist only to mine, so their whole rent counts against mining. Net/day is revenue minus rent (GPU + storage + bandwidth).
- Idle mining runs on job workers between jobs. Those machines are rented for jobs anyway, so only the share of rent matching the time spent mining (last 2h) is charged to mining, and revenue is scaled the same way.
- Break-even rent is the hourly price at which a card with the fleet's typical hashrate earns exactly its rent at today's price and yield. The watchdog replaces dedicated miners that stay below break-even for an hour.
- Pool account figures (unpaid, paid, 24h change) are real balances from the pool; per-miner figures are live estimates.
Queued Jobs by Type
| Job Type | Count |
|---|
Queue Server Types
| Server Type | Count |
|---|
Current Scaling Metrics
| Server Type | Configured Min/Max | Capacity Scaling | Live (R/B/P) | Out of Service | Active Jobs | Queued Comfy | Queued Est (min) | Headroom | Hourly Cost |
|---|
Job Queue
Pending, queued, and in-progress work.
| Select All | Job ID | Job | Server | Status | Age | Status Age | Estimated execution | Priority | Instance | Actions |
|---|
Untracked Vast Rentals
Rentals outside the tracked instance set.
| Instance | Label | GPU | Status | Hourly | Started | Host |
|---|
Jobs
Find a job instantly and review the latest outcomes.
Check job status
Recent jobs
Outcomes, runtimes, assigned instances, and output links.
| Job ID | Job | Server | Status | Age | Duration | Instance | Output / Error |
|---|
Configuration
Scaling overrides, ComfyUI registries, and dependency catalogs.
Server Scaling Configuration
Manage scaling parameters for different server types. Saved values override application defaults.
ComfyUI Type Registries
Manage comfyUIServerTypes/* and comfyUIJobTypes/* in Firestore.
Server Types
| Server Type | Enabled | Model Mode | Template | Readiness | Job Types | Actions |
|---|
Job Types
| Job Type | Enabled | Category | Scheduler | Allowed Server Types | Actions |
|---|
Edit
Dependency Manager Catalogs
Manage dependencies/* and serverTypeDependencyProfiles/* used by dependency-managed server types.
Dependencies
| Dependency ID | Kind | Type | File | Destination | Source | Size | Actions |
|---|
Server Type Dependency Profiles
| Profile ID | Server Type | Prefetch | Core | Warm | Dynamic | Actions |
|---|
Edit
Dependency Queue Viewer (Read-only)
| Item ID | Dep ID | State | Op | Priority | Attempts | Requested By | Updated | Error |
|---|
Admin Logs
Recent scaling and infrastructure decisions.
Critical Decisions (24h)
Actions, errors, and reasoning from recent automated decisions.
| Time | Action | Server Type | Instance | Reason / Error | Details |
|---|