Skip to content

CPU admission

cpu_overcommit under [host] adds a CPU budget, the slice’s CPUs times the factor, and a runner starts only when what it reserves of that fits too:

[host]
cpu_overcommit = 1.0
  • The slice’s CPUs are its AllowedCPUs, or the host’s, fewer when its CPUQuota allows less, read from systemd.
  • Under measured a runner reserves its job’s average CPU, predicted from the history as memory is, with no margin: the factor is the margin. Under caps it reserves its profile’s cpus.

A job that had a slot and memory but not CPU waits; after the same 10 or 2 minutes it goes first, by the same rule as memory. The status line says so (CPU 30 of 32 CPUs reserved; a job waits for CPU).

Averages are far under peaks, so on a host with many CPUs they admit nearly everything; on 8 CPUs (Simulating) the budget holds jobs. Start at 1.0: peaks that land together cost a short slowdown, not a kill.

Linux shares the slice’s CPUs between containers in proportion to each profile’s cpus, and no container gets more than its cpus. A low CPUWeight on the slice, such as 20, leaves the desktop its share.