pub struct Host {
pub slice: String,
pub isolation: Isolation,
pub allow_unsafe: bool,
pub poll_interval: Duration,
pub admission: Admission,
pub cpu_overcommit: Option<f64>,
pub memory_ceiling: Option<ByteSize>,
pub pooled_count: bool,
}Expand description
[host]: where runner containers run and how they are isolated.
Fields§
§slice: StringThe systemd slice every runner container is placed in; its MemoryMax is the budget.
isolation: IsolationHow runner containers are kept from the host.
allow_unsafe: boolMust be set for isolation = "unsafe-dind", so a copied config cannot turn it on silently.
poll_interval: DurationHow often GitHub is asked for queued jobs.
admission: AdmissionWhat a runner must fit in before it starts. Default caps.
cpu_overcommit: Option<f64>Set, a runner also starts only when its predicted CPU fits the slice’s CPUs times this; absent, CPU is not admitted.
memory_ceiling: Option<ByteSize>Set, the most memory the host can grow to, for a VM whose hypervisor adds memory as the guest asks for it:
admission counts what it can still grow by past MemTotal as available, and keeps the margin of this
(ADR 0084). Absent, or not above MemTotal, the host is the memory it has now.
pooled_count: boolUnder measured admission, whether MemoryMax and the host’s limit count what runs uses and keep room for its
growth (ADR 0099, 0108, 0130) rather than every runner’s worst case. Default on; off, both add every runner’s
take, the way back if a host meets what the count did not foresee.