Skip to content

Resource Isolation

In a multi-service environment, one misbehaving process can consume all RAM or CPU and trigger the kernel OOM killer — often taking down unrelated services (web-server, ssh, etc.).

Licensed plugins (isolation) address this with Linux Cgroups v2: hard memory caps and CPU quotas per program, so a runaway worker is contained without Docker.

Why it matters in production

Without limitsWith Cgroups
One leak can starve the whole hostOOM kills only the offending cgroup
Batch jobs steal CPU from APIscpu_quota throttles low-priority work
Need containers for isolationBare-metal isolation via super.toml

Configuration, kernel requirements, and metrics are documented in Resource Isolation (requires isolation plugin).