Documentation
¶
Overview ¶
Package gantry implements the gantry peer-to-peer OCI distribution cluster component. Gantry runs one agent DaemonSet per node fronting containerd's image pulls, so it is reconciled as a cluster singleton. Unlike the other components it defaults to enabled: it is deployed unless every Site explicitly opts out via spec.components.gantry.enabled=false.
The unbounded agent manages the per-node containerd wiring, so the operator installs only the Gantry workload and its Kubernetes configuration.
When ClusterCache/gantry exists, the operator switches an installation it reconciles to the Racer backend: Gantry then serves registry content through Racer and runs only where the Racer dataplane runs. The cache never installs Gantry by itself.
Index ¶
Constants ¶
const ( // RacerCacheName is the ClusterCache that switches an enabled Gantry // installation to its Racer backend. The Gantry agent always serves the // cache with this name (internal/gantry/racer.CacheName), so no other // cache name can select it. RacerCacheName = "gantry" )
Variables ¶
This section is empty.
Functions ¶
func EnabledFor ¶
func EnabledFor(site *unboundedv1alpha3.Site) bool
EnabledFor reports whether a Site enables the gantry component. Gantry defaults to enabled: only an explicit enabled=false opts a Site out.
Types ¶
type Component ¶
type Component struct{}
Component reconciles the gantry cluster singleton.
func (Component) ConditionType ¶
ConditionType implements component.ClusterComponent.
func (Component) Plan ¶ added in v0.3.0
func (c Component) Plan(ctx context.Context, env *component.Env, sites []unboundedv1alpha3.Site) (*component.Plan, component.Result, error)
Reconcile deploys the gantry cluster singleton whenever at least one Site does not opt out and keeps an existing installation reconciled when every Site opts out.
Source Files