Skip to main content

CGroups

Powered by systemd System and Service Manager.

Hierarchy​

Networks, nodes and commands are forming a Unified Cgroup Hierarchy (v2):

$ systemd-cgls
...
├─gont.slice
│ └─gont-barlow.slice
│   └─gont-barlow-h1.slice
│     └─gont-run-2797869.scope
│       └─2797869 sleep 3600

We rely on systemd's service manager to manage the hierarchy.

Freeze, Thaw, Kill​

All processes of a Cgroup can be controlled together:

cmd.Freeze()   // Suspends all processes, included forked sub-processes

cmd.Thaw()     // Resumes all processes

cmd.Teardown() // Sends a SIGKILL to all processes

This also works on the host and network levels.

By default, Close() will invoke Teardown().

Hence, guaranteeing that no lingering processes will stick around.

Resource Control​

We can control resource quota's in each level of the hierarchy using systemd resource control properties:

import sdopt "github.com/cunicu/gont/v2/options/systemd"

network, _ := gont.NewNetwork("mynet", sdopt.AllowedCPUs(0b1100))
host1, _ := network.AddHost("host1", sdopt.TasksMax(10))

cmd := host1.Command("long-running-command", sdopt.RuntimeMax(10 * time.Second))
cmd := host1.Command("memory-hungry-command", sdopt.MemoryMax(1 << 20))

See: systemd.resource-control