Documentation
¶
Index ¶
- Constants
- func ReportConfigLoadFailure(ctx context.Context, crdClient versioned.Interface, ref events.ConfigSourceRef, ...)
- type ConfigChangeHandler
- func (h *ConfigChangeHandler) EnableReinitialization()
- func (h *ConfigChangeHandler) RequestEffectiveReload()
- func (h *ConfigChangeHandler) SetEffectiveResolver(resolve func(*coreconfig.Config) (*ResolvedConfig, error))
- func (h *ConfigChangeHandler) SetInitialConfigVersion(version string)
- func (h *ConfigChangeHandler) SetInitialSnapshot(snapshot *ValidatedSnapshot)
- func (h *ConfigChangeHandler) Start(ctx context.Context) error
- type ReloadReason
- type ReloadRequest
- type ResolvedConfig
- type StatusUpdater
- type ValidatedSnapshot
Constants ¶
const ( // ComponentName is the unique identifier for this component. ComponentName = "configchange-handler" // EventBufferSize is the size of the event subscription buffer. // Moderate-volume component handling config and validation events. EventBufferSize = busevents.StandardSubscriberBuffer )
const ( // StatusUpdaterComponentName is the unique identifier for this component. StatusUpdaterComponentName = "status-updater" // StatusUpdaterEventBufferSize is the size of the event subscription buffer. StatusUpdaterEventBufferSize = busevents.StandardSubscriberBuffer )
const DefaultReinitDebounceInterval = 2 * time.Second
DefaultReinitDebounceInterval is the default time to wait after the last config change before signaling controller reinitialization. This allows rapid CRD updates to be coalesced, ensuring templates are fully rendered before reinitialization starts. Stays lenient where the per-watcher default (types.DefaultDebounceInterval) is near zero: reinit tears down and rebuilds every informer, so coalescing a burst of operator-initiated CRD edits is worth the added seconds.
Variables ¶
This section is empty.
Functions ¶
func ReportConfigLoadFailure ¶
func ReportConfigLoadFailure( ctx context.Context, crdClient versioned.Interface, ref events.ConfigSourceRef, failures []string, logger *slog.Logger, )
ReportConfigLoadFailure reports rejection before StatusUpdater starts. Reporting errors never mask the load failure; the pod stays NotReady and retries.
Types ¶
type ConfigChangeHandler ¶
type ConfigChangeHandler struct {
// contains filtered or unexported fields
}
ConfigChangeHandler coordinates configuration validation and detects config changes.
This component has two main responsibilities:
Validation Coordination: Subscribes to ConfigParsedEvent, sends ConfigValidationRequest, collects responses from validators using scatter-gather pattern, and publishes ConfigValidatedEvent or ConfigInvalidEvent. When validators cannot judge the config, it publishes ConfigValidationInconclusiveEvent and retries.
Change Detection: Subscribes to ConfigValidatedEvent and signals the controller to reinitialize via the configChangeCh channel with debouncing to coalesce rapid changes.
Debouncing behavior: When multiple CRD config changes arrive in rapid succession, the handler debounces the reinitialization signal. This ensures all pending renders complete before reinitialization starts, preventing the race condition where reinitialization cancels in-progress renders.
Architecture: This component bridges the gap between configuration parsing and validation, and between validation and controller reinitialization. It uses the scatter-gather pattern from the event bus for coordinating validation across multiple validators.
func NewConfigChangeHandler ¶
func NewConfigChangeHandler( eventBus *busevents.EventBus, logger *slog.Logger, configChangeCh chan *ReloadRequest, validators []string, debounceInterval time.Duration, ) *ConfigChangeHandler
NewConfigChangeHandler creates a new ConfigChangeHandler.
Parameters:
- eventBus: The EventBus to subscribe to and publish on
- logger: Structured logger for diagnostics
- configChangeCh: Channel to signal controller reinitialization with validated config
- validators: List of expected validator names (e.g., ["basic", "template", "jsonpath"])
- debounceInterval: Time to wait after last config change before triggering reinitialization. Use 0 for default (500ms).
Returns:
- *ConfigChangeHandler ready to start
func (*ConfigChangeHandler) EnableReinitialization ¶
func (h *ConfigChangeHandler) EnableReinitialization()
EnableReinitialization enables the reinitialization signaling mechanism.
This must be called after controller startup is complete to allow config changes to trigger reinitialization. During startup, ConfigValidatedEvents occur that should NOT trigger reinitialization. Calling this method signals that the startup phase is complete and future config changes should trigger reinitialization.
Note: CRDWatcher uses generation-based filtering, so status-only updates (which don't increment generation) never trigger ConfigValidatedEvents in the first place.
func (*ConfigChangeHandler) RequestEffectiveReload ¶
func (h *ConfigChangeHandler) RequestEffectiveReload()
RequestEffectiveReload asks the handler to restart from its accepted raw config after resolving API versions again.
func (*ConfigChangeHandler) SetEffectiveResolver ¶
func (h *ConfigChangeHandler) SetEffectiveResolver(resolve func(*coreconfig.Config) (*ResolvedConfig, error))
SetEffectiveResolver installs the effective-config transformation applied to every parsed config before validation (see the field doc). Like SetInitialConfigVersion, this must be called after construction and before the CRD watcher starts delivering events.
func (*ConfigChangeHandler) SetInitialConfigVersion ¶
func (h *ConfigChangeHandler) SetInitialConfigVersion(version string)
SetInitialConfigVersion sets the initial config version to prevent reinitialization on bootstrap ConfigValidatedEvent.
This must be called after fetching the initial config but before CRDWatcher starts. When CRDWatcher's informer triggers onAdd for the existing CRD, it publishes a ConfigValidatedEvent with the same version. Without this tracking, that event would trigger reinitialization, creating an infinite loop.
Parameters:
- version: The resourceVersion from the initial CRD fetch
func (*ConfigChangeHandler) SetInitialSnapshot ¶
func (h *ConfigChangeHandler) SetInitialSnapshot(snapshot *ValidatedSnapshot)
SetInitialSnapshot records the configuration and credentials owned by the running iteration. It must run before reinitialization is enabled.
func (*ConfigChangeHandler) Start ¶
func (h *ConfigChangeHandler) Start(ctx context.Context) error
Start begins processing events from the EventBus.
This method blocks until the context is canceled. The component is already subscribed to the EventBus (subscription happens in constructor). Returns nil on graceful shutdown.
Example:
go handler.Start(ctx)
type ReloadReason ¶
type ReloadReason uint8
ReloadReason identifies independent desired-state changes carried by a reload.
const ( ReloadReasonConfig ReloadReason = 1 << iota ReloadReasonCredentials ReloadReasonEffectiveConfig )
func (ReloadReason) Has ¶
func (r ReloadReason) Has(reason ReloadReason) bool
Has reports whether the reason set contains reason.
type ReloadRequest ¶
type ReloadRequest struct {
Snapshot *ValidatedSnapshot
Reasons ReloadReason
}
ReloadRequest carries the authoritative state for the next iteration. An effective-config reason requires re-resolving Snapshot.RawConfig; otherwise Snapshot.Config is already validated and must be used verbatim.
type ResolvedConfig ¶
type ResolvedConfig struct {
Config *coreconfig.Config
Resolution *coreconfig.Resolution
}
ResolvedConfig is the effective configuration and the discovery outcome that produced it. Validators and the next controller iteration consume this pair together.
type StatusUpdater ¶
type StatusUpdater struct {
*component.Base
*component.ReadySignal
// contains filtered or unexported fields
}
StatusUpdater updates HAProxyTemplateConfig status based on validation results.
This component subscribes to ConfigValidatedEvent, ConfigInvalidEvent, and ValidationFailedEvent, updating the HAProxyTemplateConfig CRD status to reflect validation state. Users can then see validation errors via `kubectl describe haproxytemplateconfig`.
Architecture: This is an event adapter that bridges validation events to Kubernetes API updates. It uses the generated typed client for HAProxyTemplateConfig status updates.
The component handles two types of validation: - Config validation (Stage 1): Template syntax, JSONPath expressions, etc. - HAProxy validation (Stage 4): Rendered config syntax check with haproxy -c.
func NewStatusUpdater ¶
func NewStatusUpdater( crdClient versioned.Interface, kubeClient kubernetes.Interface, eventBus *busevents.EventBus, logger *slog.Logger, ) *StatusUpdater
NewStatusUpdater creates a new StatusUpdater.
Parameters:
- crdClient: typed client for HAProxyTemplateConfig CRD status updates
- kubeClient: core client used to emit Kubernetes Events on the CRD
- eventBus: EventBus to subscribe to validation events
- logger: Structured logger for diagnostics
Returns:
- *StatusUpdater ready to start
func (*StatusUpdater) HandleEvent ¶
func (u *StatusUpdater) HandleEvent(event busevents.Event)
HandleEvent implements component.EventHandler, routing each validation event to the matching status-update handler using the loop context captured by Start.
type ValidatedSnapshot ¶
type ValidatedSnapshot struct {
RawConfig *coreconfig.Config
Config *coreconfig.Config
Resolution *coreconfig.Resolution
TemplateConfig any
ConfigVersion string
Credentials *coreconfig.Credentials
CredentialsVersion string
Sources []events.ConfigSourceRef
}
ValidatedSnapshot is the complete input accepted for one controller iteration. Config is effective; RawConfig retains candidate API versions for later discovery re-resolution.