👻 Ghostfleet run report

🖥️ 1000 nodes · 🎮 8000 GPUs · 📦 21546 pods · snapshot 2026-07-19T08:21:02Z

📋 Cluster shape

🖥️ Nodes
1000
🎮 GPUs
8000
📦 Pods
21546
🗄️ etcd DB size
550.9MB
⚡ Scheduler throughput
142.8pods/s (5m avg)
⏱️ Scheduling e2e p99
0.530s

🌐 API request p99 latency by verb

GET
22.99 ms
LIST
99.70 ms
POST
24.47 ms
PUT
24.60 ms
APPLY
no data
PATCH
24.66 ms
DELETE
24.96 ms
Table view
Verbp99
GET22.99 ms
LIST99.70 ms
POST24.47 ms
PUT24.60 ms
APPLYno data
PATCH24.66 ms
DELETE24.96 ms

SLO: mutating < 1s · read-only namespaced < 5s · read-only cluster-scoped < 30s

🗄️ etcd request p99 latency by operation

update
23.74 ms
delete
23.59 ms
create
23.42 ms
get
22.53 ms
list
20.80 ms
listWithCount
24.73 ms
Table view
Operationp99
update23.74 ms
delete23.59 ms
create23.42 ms
get22.53 ms
list20.80 ms
listWithCount24.73 ms

⚖️ Priority & Fairness

✓ No APF rejections or queue wait observed in this window.

👀 Watchers & inflight requests

No watcher data in this window.

Inflight requests
KindCount
mutating14
readOnly4

Generated from results/20260719-102054/ — raw PromQL snapshots in that folder are the source of truth.