kubernetes

package
v0.9.0 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Aug 4, 2026 License: MPL-2.0 Imports: 30 Imported by: 0

Documentation

Index

Constants

View Source 
const (
	// VersionMarkerConfigMapName is the ConfigMap, in the gitops namespace, that holds the marker.
	VersionMarkerConfigMapName = "windsor-version-marker"

	// VersionMarkerPhaseIdle marks a context with no upgrade in flight.
	VersionMarkerPhaseIdle = "idle"

	// VersionMarkerPhaseUpgrading is the phase of a context with an upgrade in flight. apply's
	// version gate refuses any non-idle phase, so an interrupted upgrade is resumed, not reconciled.
	VersionMarkerPhaseUpgrading = "upgrading"
)

Variables

This section is empty.

Functions

func SourcesEqual added in v0.9.0 

func SourcesEqual(a, b map[string]SourceRef) bool

SourcesEqual reports whether two applied-source sets are identical: the same source names, each carrying the same URL and resolved ref. It is the version-equality test the apply gate uses — equal sets mean the blueprint matches what is applied, so apply may reconcile in place rather than redirect to upgrade. SourceRef is a plain comparable struct, so values compare by field.

Types

type BaseKubernetesManager 

type BaseKubernetesManager struct {
	// contains filtered or unexported fields
}

BaseKubernetesManager implements KubernetesManager interface

func NewKubernetesManager 

func NewKubernetesManager(kubernetesClient client.KubernetesClient, configHandler config.ConfigHandler) *BaseKubernetesManager

NewKubernetesManager creates a new instance of BaseKubernetesManager. The configHandler is used to retrieve context name and context ID for CommonMetadata labels.

func (*BaseKubernetesManager) ApplyBlueprint 

func (k *BaseKubernetesManager) ApplyBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error

ApplyBlueprint applies the entire blueprint to the cluster in the proper sequence. It creates the target namespace, applies all blueprint source repositories (Git and OCI), applies all individual sources, applies any standalone ConfigMaps, and finally applies all kustomizations and their associated ConfigMaps. This orchestrates a complete blueprint installation following the intended order. Context ownership labels are stamped on each Kustomization's ObjectMeta (so the objects are selectable by context) and propagated to managed resources via CommonMetadata, using context info from the config handler. Returns an error if any step fails.

func (*BaseKubernetesManager) ApplyConfigMap 

func (k *BaseKubernetesManager) ApplyConfigMap(name, namespace string, data map[string]string) error

ApplyConfigMap creates or updates a ConfigMap using SSA

func (*BaseKubernetesManager) ApplyGitRepository 

func (k *BaseKubernetesManager) ApplyGitRepository(repo *sourcev1.GitRepository) error

ApplyGitRepository creates or updates a GitRepository resource using SSA

func (*BaseKubernetesManager) ApplyKustomization 

func (k *BaseKubernetesManager) ApplyKustomization(kustomization kustomizev1.Kustomization) error

ApplyKustomization creates or updates a Kustomization resource using SSA

func (*BaseKubernetesManager) ApplyOCIRepository 

func (k *BaseKubernetesManager) ApplyOCIRepository(repo *sourcev1.OCIRepository) error

ApplyOCIRepository creates or updates an OCIRepository resource using SSA

func (*BaseKubernetesManager) ApplySecret added in v0.9.0 

func (k *BaseKubernetesManager) ApplySecret(name, namespace string, stringData map[string]string, owner string) error

ApplySecret creates or updates a Secret using SSA. Values are supplied as plaintext in stringData (write-only; the API server folds them into data) and are never logged. Mirrors ApplyConfigMap's server-side-apply handling, including its immutable-field guard: Kubernetes rejects an update that changes Secret.type, so if an existing Secret's type differs from the newly resolved one, ApplySecret deletes it first rather than SSA-merging a rejected change. It stamps the context ownership labels plus a secret-owner label naming the kustomization the secret belongs to; that label is set only by CLI placement (never by Flux), so PruneSecrets can find and reclaim CLI-placed secrets without ever touching a Flux-managed one. The Secret's type and stringData are resolved by secretTypeAndData: stringData already carrying ".dockerconfigjson", or carrying docker-username/docker-password (docker-server optional), produces a kubernetes.io/dockerconfigjson Secret for imagePullSecrets; anything else stays Opaque.

func (*BaseKubernetesManager) ApplyVersionMarker added in v0.9.0 

func (k *BaseKubernetesManager) ApplyVersionMarker(namespace string, marker VersionMarker) error

ApplyVersionMarker writes the applied-version marker ConfigMap to the namespace, recording which blueprint version the context is running. The marker is stored as JSON in a single ConfigMap so its encoding can evolve without churning Kustomization labels.

func (*BaseKubernetesManager) CheckGitRepositoryStatus 

func (k *BaseKubernetesManager) CheckGitRepositoryStatus() error

CheckGitRepositoryStatus checks the status of all GitRepository and OCIRepository resources

func (*BaseKubernetesManager) CreateNamespace 

func (k *BaseKubernetesManager) CreateNamespace(name string) error

CreateNamespace creates a new namespace

func (*BaseKubernetesManager) DeleteBlueprint 

func (k *BaseKubernetesManager) DeleteBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error

DeleteBlueprint tears the blueprint down in two phases:

  1. Destroy-only kustomizations: apply, wait ready, delete in reverse-topological order. These are blueprint entries that exist only at destroy time for bespoke teardown work (e.g. backups, snapshots, last-mile state exports). Any errors from this phase are joined and returned immediately — the destroy walk does not start until the destroy hooks succeed.

  2. Regular kustomizations in reverse-topological order. Each Kustomization carries spec.deletionPolicy=WaitForTermination (set at apply time by ToFluxKustomization), so DELETE blocks until every managed resource is fully gone from etcd. The chain for cloud resources is:

    K8s DELETE → controller's finalizer holds the object in etcd → controller calls cloud API to release external state → finalizer lifts → object NotFound → WaitForTermination satisfied

    This is what makes CSI volumes, LB Services, Ingresses, and cert-manager Certificates clean up cloud-side without the orchestrator ever calling a cloud API. external-dns is the one outlier — it has no finalizer; the DNS record is removed on its next reconcile after the K8s object disappears.

    Phase 2 aborts on the first per-Kustomization failure (typically an inventory item stuck on a cloud-controller finalizer). Continuing the walk would tear down upstream controllers (lb-base, dns, pki-base) still needed to lift those finalizers, turning a recoverable stuck-Kustomization into a cascade of orphaned cloud resources. Re-running destroy after the operator restores the controller picks up where it left off — already-deleted Kustomizations short-circuit to NotFound on retry.

func (*BaseKubernetesManager) DeleteKustomization 

func (k *BaseKubernetesManager) DeleteKustomization(name, namespace string) error

DeleteKustomization removes a Kustomization resource using background deletion. Background deletion allows the kustomization to enter "Terminating" state while its children are deleted in the background. The method waits for the deletion to complete.

On timeout (Flux's WaitForTermination has not lifted its finalizer within kustomizationReconcileTimeout), the method returns an error rather than stripping finalizers. The timeout almost always means an inventory item is stuck on a cloud-controller finalizer (CSI external-attacher, aws-load-balancer-controller's service.k8s.aws/resources, cert-manager) — stripping the Kustomization's own finalizer at that point reaps the Kustomization from etcd but leaves the inventory items orphaned in their namespace, where the controller that should lift their finalizers may already be torn down. The orphan then blocks namespace termination and (worse) lets terraform proceed to destroy the cluster while cloud resources (LBs, EBS volumes, DNS records) leak. Returning an error here surfaces the stuck state so the operator can re-deploy the controller, let it finish cleanup, and re-run destroy — vs. silently masking the failure.

func (*BaseKubernetesManager) DeleteNamespace 

func (k *BaseKubernetesManager) DeleteNamespace(name string) error

DeleteNamespace deletes the specified namespace using foreground deletion. Foreground deletion ensures all resources in the namespace are removed before the namespace is deleted. This method waits for the deletion to complete before returning. Returns nil if the namespace is deleted successfully, or an error if deletion fails or times out.

func (*BaseKubernetesManager) GetHelmReleasesForKustomization 

func (k *BaseKubernetesManager) GetHelmReleasesForKustomization(name, namespace string) ([]helmv2.HelmRelease, error)

GetHelmReleasesForKustomization gets HelmReleases associated with a Kustomization

func (*BaseKubernetesManager) GetKustomizationInventory added in v0.9.0 

func (k *BaseKubernetesManager) GetKustomizationInventory(name, namespace string) ([]InventoryEntry, error)

GetKustomizationInventory returns the list of resources Flux is currently tracking for the named Kustomization, decoded from its .status.inventory.entries field. This is what flux uses to drive prune behavior, so it is the authoritative source for "what will be deleted when this Kustomization is removed." Returns (nil, nil) when the Kustomization itself is absent (a destroy-plan caller should treat that as "not deployed" rather than an error). Returns an empty slice when the Kustomization exists but has no inventory yet (e.g., suspended, or never reconciled). API errors reading the Kustomization or its inventory propagate. Individual entries that fail to decode (malformed IDs, unexpected field shapes) are silently dropped — flux always emits well-formed IDs, so this branch is rare in practice, and resilience matters more than completeness here: failing the whole destroy preview because of one corrupt entry would be worse than rendering a slightly truncated list.

func (*BaseKubernetesManager) GetKustomizationReadiness added in v0.9.0 

func (k *BaseKubernetesManager) GetKustomizationReadiness(names []string) (map[string]bool, error)

GetKustomizationReadiness returns whether each named Kustomization currently reports Ready=True, in the gitops namespace. Unlike GetKustomizationStatus it never fails on a Kustomization in a failed state — a failed one is simply reported not-ready — so a convergence driver can keep nudging it toward Ready rather than aborting. Names absent from the cluster report false; only an API list error propagates.

func (*BaseKubernetesManager) GetKustomizationStatus 

func (k *BaseKubernetesManager) GetKustomizationStatus(names []string) (map[string]bool, error)

GetKustomizationStatus returns a map indicating readiness for each specified kustomization in the default Flux system namespace. If a kustomization is not found, its status is set to false. If any kustomization has a Ready condition with Status False and Reason "ReconciliationFailed", an error is returned with the failure message.

func (*BaseKubernetesManager) GetNodeReadyStatus 

func (k *BaseKubernetesManager) GetNodeReadyStatus(ctx context.Context, nodeNames []string) (map[string]bool, error)

GetNodeReadyStatus returns a map of node names to their Ready condition status. Returns a map of node names to Ready status (true if Ready, false if NotReady), or an error if listing fails.

func (*BaseKubernetesManager) GetVersionMarker added in v0.9.0 

func (k *BaseKubernetesManager) GetVersionMarker(namespace string) (VersionMarker, bool, error)

GetVersionMarker reads the applied-version marker ConfigMap from the namespace, reporting false when no marker is present — a missing ConfigMap (pre-bootstrap context) or one without marker data (legacy cluster). It returns an error only on a real read or decode failure, so callers can tell "no marker yet" (proceed as legacy) apart from "could not read the marker" (cluster unreachable).

func (*BaseKubernetesManager) KustomizationExists added in v0.9.0 

func (k *BaseKubernetesManager) KustomizationExists(name, namespace string) (bool, error)

KustomizationExists returns true if a Kustomization resource with the given name exists in the given namespace. Returns false (not an error) when the resource is simply absent; propagates other API errors.

func (*BaseKubernetesManager) ListPrunableKustomizations added in v0.9.0 

func (k *BaseKubernetesManager) ListPrunableKustomizations(blueprint *blueprintv1alpha1.Blueprint, namespace string) ([]string, error)

ListPrunableKustomizations returns the names of this context's Kustomizations that the blueprint no longer declares — exactly what PruneBlueprint would delete, in reverse-dependency order. It is the read-only input to plan's prune preview and upgrade's confirmation gate; it deletes nothing.

func (*BaseKubernetesManager) NamespaceExists added in v0.9.0 

func (k *BaseKubernetesManager) NamespaceExists(name string) (bool, error)

NamespaceExists reports whether the named namespace exists in the cluster. Namespaces are cluster-scoped, so the lookup passes an empty namespace to GetResource. A NotFound is reported as (false, nil); any other API error propagates.

func (*BaseKubernetesManager) PruneBlueprint added in v0.9.0 

func (k *BaseKubernetesManager) PruneBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error

PruneBlueprint deletes Kustomizations belonging to the current context that are no longer part of the blueprint. It scopes strictly to this context — only objects carrying the windsorcli.dev/context-id label for this context are considered, so kustomizations owned by other contexts (or by no Windsor context) are never touched. Every non-DestroyOnly kustomization in the blueprint is treated as desired; the live remainder is deleted in reverse-dependency order (read from each object's live spec.dependsOn) so dependents tear down before their dependencies, each honoring its own deletionPolicy. The caller passes the same prepared blueprint Install applied (CRD layers included) so the synthesized crds/crds-<source> layers are recognized as desired and not pruned. Deletion errors are collected and joined rather than aborting on the first.

func (*BaseKubernetesManager) PruneSecrets added in v0.9.0 

func (k *BaseKubernetesManager) PruneSecrets(desired map[string]map[string]bool) error

PruneSecrets deletes the CLI-placed secrets for this context that the latest placement no longer wants, reconciling the cluster to the desired set. desired maps a namespace to the set of secret names just placed there; a secret whose (namespace, name) is absent is deleted. It lists only secrets bearing this context's id and the secret-owner marker — a label set solely by ApplySecret, never by Flux — so it reclaims exactly what the CLI placed (a secret dropped from a fan-out list, a secret removed from a system, or every CLI secret when desired is empty) and never a Flux-managed secret that merely inherited the context labels via CommonMetadata. It fails closed when the context id is unset or is not a valid label value, since without a well-formed id pruning cannot be scoped to this context — a malformed id would otherwise build a selector the API server rejects, failing every placement with an opaque error rather than a message that names the bad id.

func (*BaseKubernetesManager) RollWorkloadsForSecret added in v0.9.0 

func (k *BaseKubernetesManager) RollWorkloadsForSecret(ctx context.Context, namespace, secretName, digest string) error

RollWorkloadsForSecret rolls the workloads in a namespace that consume the named Secret so they pick up new content, the way Kubernetes only ever rolls on a pod-template change. Because the CLI holds the resolved plaintext, it passes a precomputed content digest and stamps it as a pod-template annotation (checksum.windsorcli.dev/<secret>) on every Deployment, StatefulSet, and DaemonSet whose pod spec references the Secret via envFrom, a secretKeyRef, or a secret volume — including init containers. The digest is one-way and never surfaced beyond the namespace, so a reader of the annotation learns nothing the Secret's own RBAC did not already grant. A workload already carrying the digest is left untouched (idempotent, so unchanged content does not churn pods); a workload that does not reference the Secret is never patched. Finding no consumers is not an error — on a first apply the workload is created later by Flux and reads the Secret fresh — so this returns an error only on an API failure. The caller's context bounds the patch calls so a slow API server cannot outlast its deadline.

func (*BaseKubernetesManager) WaitForKubernetesHealthy 

func (k *BaseKubernetesManager) WaitForKubernetesHealthy(ctx context.Context, endpoint string, outputFunc func(string), nodeNames ...string) error

WaitForKubernetesHealthy waits for the Kubernetes API to become healthy within the context deadline. If nodeNames are provided, verifies all specified nodes reach Ready state before returning. Returns an error if the API is unreachable or any specified nodes are not Ready within the deadline.

func (*BaseKubernetesManager) WaitForKustomizations 

func (k *BaseKubernetesManager) WaitForKustomizations(ctx context.Context, message string, blueprint *blueprintv1alpha1.Blueprint) error

WaitForKustomizations waits for kustomizations to be ready, calculating the timeout from the longest dependency chain in the blueprint. Outputs a debug message describing the total wait timeout being used before beginning polling. The wait also honors ctx: a cancelled or deadline-exceeded context ends the wait immediately and returns ctx.Err(), so a parent SIGTERM/Ctrl+C or command deadline can interrupt it (rather than requiring a SIGKILL that bypasses the caller's defer cleanup). A not-found GetResource error is treated as not-ready-yet and keeps polling. Any other error (auth, connection, RBAC) counts toward kustomizationWaitMaxConsecutiveErrors: a single transient blip (a brief apiserver restart, a load-balancer failover) is tolerated and retried on the next tick, but a persistent failure like broken cluster auth ends the wait immediately with the underlying error surfaced, rather than silently retrying it for the full timeout budget.

type InventoryEntry added in v0.9.0 

type InventoryEntry struct {
	Group     string
	Kind      string
	Namespace string
	Name      string
}

InventoryEntry identifies one resource Flux is tracking for a Kustomization, decoded from a single .status.inventory.entries[] record. The encoded form flux writes is "<namespace>_<name>_<group>_<kind>"; namespace is empty for cluster-scoped resources, group is empty for core API objects ("v1"). These entries are exactly what flux deletes when a Kustomization is removed, so they are the truthful source for "what will go away on destroy."

type KubernetesManager 

type KubernetesManager interface {
	ApplyKustomization(kustomization kustomizev1.Kustomization) error
	DeleteKustomization(name, namespace string) error
	WaitForKustomizations(ctx context.Context, message string, blueprint *blueprintv1alpha1.Blueprint) error
	CreateNamespace(name string) error
	DeleteNamespace(name string) error
	ApplyConfigMap(name, namespace string, data map[string]string) error
	ApplySecret(name, namespace string, stringData map[string]string, owner string) error
	PruneSecrets(desired map[string]map[string]bool) error
	RollWorkloadsForSecret(ctx context.Context, namespace, secretName, digest string) error
	GetHelmReleasesForKustomization(name, namespace string) ([]helmv2.HelmRelease, error)
	ApplyGitRepository(repo *sourcev1.GitRepository) error
	ApplyOCIRepository(repo *sourcev1.OCIRepository) error
	CheckGitRepositoryStatus() error
	GetKustomizationStatus(names []string) (map[string]bool, error)
	GetKustomizationReadiness(names []string) (map[string]bool, error)
	KustomizationExists(name, namespace string) (bool, error)
	NamespaceExists(name string) (bool, error)
	GetKustomizationInventory(name, namespace string) ([]InventoryEntry, error)
	WaitForKubernetesHealthy(ctx context.Context, endpoint string, outputFunc func(string), nodeNames ...string) error
	GetNodeReadyStatus(ctx context.Context, nodeNames []string) (map[string]bool, error)
	ApplyBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error
	DeleteBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error
	PruneBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error
	ListPrunableKustomizations(blueprint *blueprintv1alpha1.Blueprint, namespace string) ([]string, error)
	ApplyVersionMarker(namespace string, marker VersionMarker) error
	GetVersionMarker(namespace string) (VersionMarker, bool, error)
}

KubernetesManager defines methods for Kubernetes resource management

type MockKubernetesManager 

type MockKubernetesManager struct {
	ApplyKustomizationFunc              func(kustomization kustomizev1.Kustomization) error
	DeleteKustomizationFunc             func(name, namespace string) error
	WaitForKustomizationsFunc           func(ctx context.Context, message string, blueprint *blueprintv1alpha1.Blueprint) error
	GetKustomizationStatusFunc          func(names []string) (map[string]bool, error)
	GetKustomizationReadinessFunc       func(names []string) (map[string]bool, error)
	CreateNamespaceFunc                 func(name string) error
	DeleteNamespaceFunc                 func(name string) error
	ApplyConfigMapFunc                  func(name, namespace string, data map[string]string) error
	ApplySecretFunc                     func(name, namespace string, stringData map[string]string, owner string) error
	PruneSecretsFunc                    func(desired map[string]map[string]bool) error
	RollWorkloadsForSecretFunc          func(ctx context.Context, namespace, secretName, digest string) error
	ApplyVersionMarkerFunc              func(namespace string, marker VersionMarker) error
	GetVersionMarkerFunc                func(namespace string) (VersionMarker, bool, error)
	GetHelmReleasesForKustomizationFunc func(name, namespace string) ([]helmv2.HelmRelease, error)
	ApplyGitRepositoryFunc              func(repo *sourcev1.GitRepository) error
	ApplyOCIRepositoryFunc              func(repo *sourcev1.OCIRepository) error
	CheckGitRepositoryStatusFunc        func() error
	KustomizationExistsFunc             func(name, namespace string) (bool, error)
	NamespaceExistsFunc                 func(name string) (bool, error)
	GetKustomizationInventoryFunc       func(name, namespace string) ([]InventoryEntry, error)
	WaitForKubernetesHealthyFunc        func(ctx context.Context, endpoint string, outputFunc func(string), nodeNames ...string) error
	GetNodeReadyStatusFunc              func(ctx context.Context, nodeNames []string) (map[string]bool, error)
	ApplyBlueprintFunc                  func(blueprint *blueprintv1alpha1.Blueprint, namespace string) error
	DeleteBlueprintFunc                 func(blueprint *blueprintv1alpha1.Blueprint, namespace string) error
	PruneBlueprintFunc                  func(blueprint *blueprintv1alpha1.Blueprint, namespace string) error
	ListPrunableKustomizationsFunc      func(blueprint *blueprintv1alpha1.Blueprint, namespace string) ([]string, error)
}

MockKubernetesManager is a mock implementation of KubernetesManager interface for testing

func NewMockKubernetesManager 

func NewMockKubernetesManager() *MockKubernetesManager

NewMockKubernetesManager creates a new instance of MockKubernetesManager

func (*MockKubernetesManager) ApplyBlueprint 

func (m *MockKubernetesManager) ApplyBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error

ApplyBlueprint implements KubernetesManager interface

func (*MockKubernetesManager) ApplyConfigMap 

func (m *MockKubernetesManager) ApplyConfigMap(name, namespace string, data map[string]string) error

ApplyConfigMap implements KubernetesManager interface

func (*MockKubernetesManager) ApplyGitRepository 

func (m *MockKubernetesManager) ApplyGitRepository(repo *sourcev1.GitRepository) error

ApplyGitRepository implements KubernetesManager interface

func (*MockKubernetesManager) ApplyKustomization 

func (m *MockKubernetesManager) ApplyKustomization(kustomization kustomizev1.Kustomization) error

ApplyKustomization implements KubernetesManager interface

func (*MockKubernetesManager) ApplyOCIRepository 

func (m *MockKubernetesManager) ApplyOCIRepository(repo *sourcev1.OCIRepository) error

ApplyOCIRepository implements KubernetesManager interface

func (*MockKubernetesManager) ApplySecret added in v0.9.0 

func (m *MockKubernetesManager) ApplySecret(name, namespace string, stringData map[string]string, owner string) error

ApplySecret implements KubernetesManager interface

func (*MockKubernetesManager) ApplyVersionMarker added in v0.9.0 

func (m *MockKubernetesManager) ApplyVersionMarker(namespace string, marker VersionMarker) error

ApplyVersionMarker implements KubernetesManager interface

func (*MockKubernetesManager) CheckGitRepositoryStatus 

func (m *MockKubernetesManager) CheckGitRepositoryStatus() error

CheckGitRepositoryStatus checks the status of all GitRepository resources

func (*MockKubernetesManager) CreateNamespace 

func (m *MockKubernetesManager) CreateNamespace(name string) error

CreateNamespace implements KubernetesManager interface

func (*MockKubernetesManager) DeleteBlueprint 

func (m *MockKubernetesManager) DeleteBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error

DeleteBlueprint implements KubernetesManager interface

func (*MockKubernetesManager) DeleteKustomization 

func (m *MockKubernetesManager) DeleteKustomization(name, namespace string) error

DeleteKustomization implements KubernetesManager interface

func (*MockKubernetesManager) DeleteNamespace 

func (m *MockKubernetesManager) DeleteNamespace(name string) error

DeleteNamespace implements KubernetesManager interface

func (*MockKubernetesManager) GetHelmReleasesForKustomization 

func (m *MockKubernetesManager) GetHelmReleasesForKustomization(name, namespace string) ([]helmv2.HelmRelease, error)

GetHelmReleasesForKustomization implements KubernetesManager interface

func (*MockKubernetesManager) GetKustomizationInventory added in v0.9.0 

func (m *MockKubernetesManager) GetKustomizationInventory(name, namespace string) ([]InventoryEntry, error)

GetKustomizationInventory implements KubernetesManager interface

func (*MockKubernetesManager) GetKustomizationReadiness added in v0.9.0 

func (m *MockKubernetesManager) GetKustomizationReadiness(names []string) (map[string]bool, error)

GetKustomizationReadiness implements KubernetesManager interface. The default reports every requested kustomization Ready so a convergence pass returns immediately in tests that do not exercise it; tests that drive convergence set GetKustomizationReadinessFunc.

func (*MockKubernetesManager) GetKustomizationStatus 

func (m *MockKubernetesManager) GetKustomizationStatus(names []string) (map[string]bool, error)

GetKustomizationStatus implements KubernetesManager interface

func (*MockKubernetesManager) GetNodeReadyStatus 

func (m *MockKubernetesManager) GetNodeReadyStatus(ctx context.Context, nodeNames []string) (map[string]bool, error)

GetNodeReadyStatus returns a map of node names to their Ready condition status.

func (*MockKubernetesManager) GetVersionMarker added in v0.9.0 

func (m *MockKubernetesManager) GetVersionMarker(namespace string) (VersionMarker, bool, error)

GetVersionMarker implements KubernetesManager interface

func (*MockKubernetesManager) KustomizationExists added in v0.9.0 

func (m *MockKubernetesManager) KustomizationExists(name, namespace string) (bool, error)

KustomizationExists implements KubernetesManager interface

func (*MockKubernetesManager) ListPrunableKustomizations added in v0.9.0 

func (m *MockKubernetesManager) ListPrunableKustomizations(blueprint *blueprintv1alpha1.Blueprint, namespace string) ([]string, error)

ListPrunableKustomizations implements KubernetesManager interface

func (*MockKubernetesManager) NamespaceExists added in v0.9.0 

func (m *MockKubernetesManager) NamespaceExists(name string) (bool, error)

NamespaceExists implements KubernetesManager interface

func (*MockKubernetesManager) PruneBlueprint added in v0.9.0 

func (m *MockKubernetesManager) PruneBlueprint(blueprint *blueprintv1alpha1.Blueprint, namespace string) error

PruneBlueprint implements KubernetesManager interface

func (*MockKubernetesManager) PruneSecrets added in v0.9.0 

func (m *MockKubernetesManager) PruneSecrets(desired map[string]map[string]bool) error

PruneSecrets implements KubernetesManager interface

func (*MockKubernetesManager) RollWorkloadsForSecret added in v0.9.0 

func (m *MockKubernetesManager) RollWorkloadsForSecret(ctx context.Context, namespace, secretName, digest string) error

RollWorkloadsForSecret implements KubernetesManager interface

func (*MockKubernetesManager) WaitForKubernetesHealthy 

func (m *MockKubernetesManager) WaitForKubernetesHealthy(ctx context.Context, endpoint string, outputFunc func(string), nodeNames ...string) error

WaitForKubernetesHealthy waits for the Kubernetes API endpoint to be healthy with polling and timeout

func (*MockKubernetesManager) WaitForKustomizations 

func (m *MockKubernetesManager) WaitForKustomizations(ctx context.Context, message string, blueprint *blueprintv1alpha1.Blueprint) error

WaitForKustomizations implements KubernetesManager interface

type Shims 

type Shims struct {
	// Other operations
	RegexpMatchString func(pattern string, s string) (bool, error)
	YamlMarshal       func(v any) ([]byte, error)
	YamlUnmarshal     func(data []byte, v any, opts ...yaml.JSONOpt) error
	K8sYamlUnmarshal  func(data []byte, v any, opts ...yaml.JSONOpt) error
	TimeSleep         func(d time.Duration)
	ToUnstructured    func(obj any) (map[string]any, error)
	FromUnstructured  func(obj map[string]any, target any) error
}

Shims provides testable interfaces for external dependencies

func NewShims 

func NewShims() *Shims

NewShims creates a new Shims instance with default implementations

type SourceRef added in v0.9.0 

type SourceRef struct {
	URL string `json:"url,omitempty"`
	Ref string `json:"ref,omitempty"`
}

SourceRef records one blueprint source as applied: its URL and resolved human reference (tag/branch/semver/commit). The resolved digest is recorded later, when the version gate that consumes it is built.

type VersionMarker added in v0.9.0 

type VersionMarker struct {
	SchemaVersion  int                  `json:"schemaVersion"`
	Phase          string               `json:"phase"`
	AppliedSources map[string]SourceRef `json:"appliedSources,omitempty"`
	TargetSources  map[string]SourceRef `json:"targetSources,omitempty"`
}

VersionMarker is the authoritative record of the blueprint version a context is running. Because a context can carry several independently-versioned sources, the "version" is the set of source refs, not a single string. Phase tracks an in-flight upgrade (idle for a settled context); the target source set is recorded by upgrade and is absent at bootstrap.

func BuildTransitionMarker added in v0.9.0 

func BuildTransitionMarker(applied map[string]SourceRef, blueprint *blueprintv1alpha1.Blueprint) (VersionMarker, error)

BuildTransitionMarker returns an in-flight marker for an upgrade toward blueprint: the applied source set (empty for a legacy context) plus the blueprint's set as the target, under the upgrading phase. WriteVersionMarker replaces it with a settled marker once the upgrade succeeds.

func BuildVersionMarker added in v0.9.0 

func BuildVersionMarker(blueprint *blueprintv1alpha1.Blueprint) (VersionMarker, error)

BuildVersionMarker captures a settled (idle) marker from the applied blueprint: the repository source (keyed by the blueprint name) and every declared remote source, each with its resolved reference. Local template sources are skipped — they track the working tree and have no version. It errors on a source-name collision (two sources sharing a name, or a source matching the repository's blueprint name) rather than silently overwriting and misrepresenting what was applied.

func ParseVersionMarker added in v0.9.0 

func ParseVersionMarker(data map[string]string) (VersionMarker, bool, error)

ParseVersionMarker decodes a marker from ConfigMap data, reporting false when no marker is present.

func (VersionMarker) ToConfigMapData added in v0.9.0 

func (m VersionMarker) ToConfigMapData() (map[string]string, error)

ToConfigMapData encodes the marker as ConfigMap data (a single JSON document).

Source Files

View all Source files

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.