Dynamic limits
Per-user limits are applied based on workload pressure and operating conditions.
Product · Linux Resource Control
Govern Linux resources before saturation becomes an incident.
ResMan applies dynamic policy to CPU, memory and I/O using real kernel signals, with native observability and no lock-in.
Control based on real behaviour
Measurable pressure
Respond to Linux Pressure Stall Information signals.
Observability
Prometheus metrics and decisions operators can understand.
Decision
Resource contention must be managed before it becomes an incident.
On shared Linux systems, CPU, memory and I/O are contested by different workloads. When pressure persists, more processes are left waiting, operation latency rises and the load generated by one user can degrade service for everyone else. ResMan applies per-user policy only when intervention is justified: it introduces limits dynamically, keeps them in place while pressure remains significant and removes them once the system has returned to stable operating conditions. It therefore sits between static resource allocation and manual incident response, preserving operating capacity while making every intervention observable and understandable to administrators.
- Kernel-reported CPU and I/O pressure
- Time windows against false positives
- Minimum limit duration
- Release based on the user’s real load
Operational features
CPU, RAM and I/O
Dynamic resource control through cgroups v2.
Per-user policy
Individual monitoring with inclusion and exclusion filters.
History and observability
Prometheus metrics, SQLite history and a Grafana dashboard.
Dynamic configuration
Hot reload for operational settings.
Integrated notifications
Scripts and webhooks when a user is limited.
Verifiable operation
Control cycles and decisions operators can inspect.
Next step