Docs

Observability

Infrastructure signals, training performance and automatic recovery.

Avg GPU utilization
86.4%
Across active nodes
Avg GPU temperature
72°C
p99: 81°C
Power draw
196 kW
610 W / GPU average
Active alerts
3
1 critical · 2 warning

GPU utilization by node

%

GPU temperature & power

°C / W

Node network throughput

GB/s

Training jobs & infrastructure events

tokens/s

Alert rules

RuleStatus

Telemetry forwarding