Nvidia extends Mission Control recovery to Kubernetes workloads

Nvidia says Mission Control 2.5 is available today, extending its automatic failure detection and recovery service to Kubernetes workloads managed with Nvidia Run:ai. The service previously supported high-performance computing workloads using Slurm. The software separates job recovery from hardware repair. It uses job context and failure history to guide restarts, retries or rescheduling, while a hardware service can isolate suspect nodes and run diagnostics. Workloads can therefore resume without waiting for hardware root-cause analysis. Transient job failures can trigger retries without removing healthy hardware from the scheduling pool. Recovery actions follow scheduler policy; the release's new AI-enabled network anomaly detection service is beta only.