prepareSyncFlow computes the required state for synchronizing PodGang resources.

GetAllOperators returns all components operators.

WithPodCliqueParameters is a convenience function that creates a PodCliqueTemplateSpec given the parameters and adds it to the PodCliqueSet.

---------------------------- Helper Functions ---------------------------- defaultTASConfig returns a default TAS configuration with TAS disabled. Thi

handleErrorAndExit gracefully handles errors before exiting the program.

newPCS returns a minimal PCS with one clique template ("worker") that has the given ResourceSharing specs and ResourceClaimTemplates.

--- Convenience methods that delegate to managers with suite-scoped defaults --- newPodManager creates a PodManager for internal use by convenience me

newResourceType creates a resourceType from a GroupVersionKind. Resource (plural) is derived from Kind, display name is Kind + "s".

waitForComputeDomainCount waits for the specified number of ComputeDomains for a PCS

ApplyYAMLFile applies a YAML file to the cluster.

Close tears down the port-forward. Safe to call multiple times.

DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.

GetDefault returns the backend marked as default in OperatorConfiguration (scheduler.defaultProfileName).

New creates a new ComputeDomain operator for managing ComputeDomain resources within PodCliqueSets.

RecordErrorForObjects records an error for a specific client.Client method and object keys.

VerifyAllPodsArePendingWithSleep verifies all pods are pending after a fixed delay.

WithLogger sets the logger for diagnostic messages during polling.

applyMNNVLYAML reads a YAML fixture, overrides the PCS metadata.name to the given pcsName (so every test gets a unique resource), and applies it.

buildTestPCS constructs a PodCliqueSet for testing based on provided parameters. If cliques is nil or empty, a default worker clique will be added.

captureExistingPodNames returns a map of all current pod names for the workload.

createManagerOptions constructs controller-runtime Manager options from operator configuration.

createOrUpdatePodGangs creates or updates all expected PodGangs. PodGangs are created with empty podReferences, Initialized=False. Once all pods are c

createPCSWithGPU creates a PCS with GPU requirements and specified replicas

newTestPod creates a pod with the given name, template hash label, and options applied.

New creates a new Service operator for managing Service resources within PodCliqueSets

NewHandler creates a new handler for PodCliqueSet Webhook. It reads TopologyAwareScheduling, Network, and Scheduler from the operator configuration. o

ScalePCSGInstanceAndWait scales a specific PCSG instance and waits for expected pod conditions.

UncordonNodesAndWaitForPods uncordons nodes and waits for pods to be ready.

WithAnnotations sets the annotations for the PodCliqueSet.

WithCliqueNames sets the CliqueNames field for the PodCliqueScalingGroup.

WithContainer adds a container to the PodSpec.

deletePCS deletes a PCS by name

hasErrors returns true if any errors occurred during sync.

Load a SetupConfig from a YAML file with optional overlay YAMLs and --set overrides. Precedence (low → high): 1. Base YAML (base_path)

Test helper functions

Run a kubectl command via subprocess and return (success, stdout, stderr). Uses subprocess instead of sh because kubectl output parsing requires

FetchFunc returns a FetchFunc that lists pods in the given namespace with the given label selector.

GetErrors returns the errors from the reconcile step.

GetExpectations gets the recorded expectations against a controlleeKey.

New creates a Volcano scheduler backend from the given scheduler profile.

New creates an instance of RoleBinding components operator.

New creates an instance of Secret components operator.

New creates an instance of Role components operator.

New creates an instance of ServiceAccount components operator.

ScalePCS scales a PodCliqueSet to the specified replica count.

ScalePCSAndWait scales a PCS and waits for the expected pod conditions.

createValidPCSWithPCSGConfigAnnotations creates a fully valid PCS with a PCSG config carrying the given annotations, for testing spec-level validation

---------------------------- Helper Functions ---------------------------- getDefaultTASConfig returns a default TAS configuration with TAS disabled.

pclqResourceClaimLabels returns the standard RC labels plus grove.io/podclique so that PCLQ-level RCs can be listed by a single label selector.

Run a command, handling errors gracefully. Returns (exit_code, output). Pass _ok_code=[0, 1, ...] to suppress exceptions for expected exit codes.

sliceMustHaveUniqueElements validates that all elements in a string slice are unique.

waitForPCLQ waits for a PCLQ to exist and returns it

waitForRollingUpdateComplete waits for rolling update to complete by checking UpdatedReplicas. Uses tc.Workload.Name as the PCS name and tc.Timeout fo

GetExistingPodGangs fetches all existing PodGangs that are managed by Grove in the given namespace.

NewOperatorRegistry creates a new OperatorRegistry.

ResourceSharersFromPCLQ converts PCLQ-level specs to the common ResourceSharer interface.

SyncTopology creates or updates the scheduler-specific topology resource for the given ClusterTopologyBinding. Called for backends not listed in the C

Wait blocks until all async after-phase hooks have completed. Call this after Run() returns to ensure no background goroutines are truncated.

WithCondition adds a condition to the Pod's status.

WithPodCliqueScalingGroupConfig adds a PodCliqueScalingGroupConfig to the PodCliqueSet.

WithTopologyConstraint sets the TopologyConstraint for the PodCliqueSet template.

Test helpers

createPCSWithMNNVLEnabled creates a PCS with MNNVL enabled via mnnvl-group annotation

createScalingGroupConfig creates a basic PodCliqueScalingGroupConfig for testing.

effectiveQueueFromAnnotations returns the effective Volcano queue for the given annotations. If the queue annotation is missing or empty, it defaults

getDefaultNetworkConfig returns a default network configuration with MNNVL disabled.

Run a command suppressing stdout and stderr. Returns the result.

---------------------------- validate update of PodCliqueSet ----------------------------------------------- validateUpdate validates the update to a

verifyRCState waits for the expected RC count and verifies exact RC names.

AddAndPatchFinalizer uses merge-patch strategy to patch the given object with the given finalizer.

CloseQuietly safely closes an io.Closer, ignoring and suppressing any error during the close operation.

DeleteClusterTopology deletes a ClusterTopologyBinding CR by name.

DownloadForPhase downloads pprof profiles for a completed phase in parallel: - CPU profile spanning [start, end] (full phase window) - Memory snapshot

ManageWebhookCerts manages webhook certificates based on the CertProvisionMode configuration. When mode=auto: uses cert-controller for automatic certi

NewTimelineTracker constructs a new timeline tracker with required metadata.