@@ -68,8 +68,6 @@ type ManagerImpl struct {
6868
6969// NewManagerImpl creates a new manager.
7070func NewManagerImpl (kubeClient clientset.Interface , stateFileDirectory string , nodeName types.NodeName ) (* ManagerImpl , error ) {
71- klog .V (2 ).InfoS ("Creating DRA manager" )
72-
7371 claimInfoCache , err := newClaimInfoCache (stateFileDirectory , draManagerStateFileName )
7472 if err != nil {
7573 return nil , fmt .Errorf ("failed to create claimInfo cache: %+v" , err )
@@ -91,15 +89,16 @@ func NewManagerImpl(kubeClient clientset.Interface, stateFileDirectory string, n
9189}
9290
9391// Start starts the reconcile loop of the manager.
94- func (m * ManagerImpl ) Start (activePods ActivePodsFunc , sourcesReady config.SourcesReady ) error {
92+ func (m * ManagerImpl ) Start (ctx context. Context , activePods ActivePodsFunc , sourcesReady config.SourcesReady ) error {
9593 m .activePods = activePods
9694 m .sourcesReady = sourcesReady
97- go wait .Until ( func () { m .reconcileLoop () }, m .reconcilePeriod , wait . NeverStop )
95+ go wait .UntilWithContext ( ctx , func (ctx context. Context ) { m .reconcileLoop (ctx ) }, m .reconcilePeriod )
9896 return nil
9997}
10098
10199// reconcileLoop ensures that any stale state in the manager's claimInfoCache gets periodically reconciled.
102- func (m * ManagerImpl ) reconcileLoop () {
100+ func (m * ManagerImpl ) reconcileLoop (ctx context.Context ) {
101+ logger := klog .FromContext (ctx )
103102 // Only once all sources are ready do we attempt to reconcile.
104103 // This ensures that the call to m.activePods() below will succeed with
105104 // the actual active pods list.
@@ -140,8 +139,8 @@ func (m *ManagerImpl) reconcileLoop() {
140139
141140 // Loop through all inactive pods and call UnprepareResources on them.
142141 for _ , podClaims := range inactivePodClaims {
143- if err := m .unprepareResources (podClaims .uid , podClaims .namespace , podClaims .claimNames ); err != nil {
144- klog . ErrorS ( err , "Unpreparing pod resources in reconcile loop" , "podUID" , podClaims .uid )
142+ if err := m .unprepareResources (ctx , podClaims .uid , podClaims .namespace , podClaims .claimNames ); err != nil {
143+ logger . Info ( "Unpreparing pod resources in reconcile loop failed, will retry " , "podUID" , podClaims .uid , "err" , err )
145144 }
146145 }
147146}
@@ -150,25 +149,26 @@ func (m *ManagerImpl) reconcileLoop() {
150149// for the input container, issue NodePrepareResources rpc requests
151150// for each new resource requirement, process their responses and update the cached
152151// containerResources on success.
153- func (m * ManagerImpl ) PrepareResources (pod * v1.Pod ) error {
152+ func (m * ManagerImpl ) PrepareResources (ctx context.Context , pod * v1.Pod ) error {
153+ logger := klog .FromContext (ctx )
154154 batches := make (map [string ][]* drapb.Claim )
155155 resourceClaims := make (map [types.UID ]* resourceapi.ResourceClaim )
156156 for i := range pod .Spec .ResourceClaims {
157157 podClaim := & pod .Spec .ResourceClaims [i ]
158- klog .V (3 ).InfoS ("Processing resource" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name )
158+ logger .V (3 ).Info ("Processing resource" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name )
159159 claimName , mustCheckOwner , err := resourceclaim .Name (pod , podClaim )
160160 if err != nil {
161161 return fmt .Errorf ("prepare resource claim: %v" , err )
162162 }
163163
164164 if claimName == nil {
165165 // Nothing to do.
166- klog .V (5 ).InfoS ("No need to prepare resources, no claim generated" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name )
166+ logger .V (5 ).Info ("No need to prepare resources, no claim generated" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name )
167167 continue
168168 }
169169 // Query claim object from the API server
170170 resourceClaim , err := m .kubeClient .ResourceV1alpha3 ().ResourceClaims (pod .Namespace ).Get (
171- context . TODO () ,
171+ ctx ,
172172 * claimName ,
173173 metav1.GetOptions {})
174174 if err != nil {
@@ -198,9 +198,9 @@ func (m *ManagerImpl) PrepareResources(pod *v1.Pod) error {
198198 return fmt .Errorf ("claim %s: %w" , klog .KObj (resourceClaim ), err )
199199 }
200200 claimInfo = m .cache .add (ci )
201- klog .V (6 ).InfoS ("Created new claim info cache entry" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name , "claim" , klog .KObj (resourceClaim ), "claimInfoEntry" , claimInfo )
201+ logger .V (6 ).Info ("Created new claim info cache entry" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name , "claim" , klog .KObj (resourceClaim ), "claimInfoEntry" , claimInfo )
202202 } else {
203- klog .V (6 ).InfoS ("Found existing claim info cache entry" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name , "claim" , klog .KObj (resourceClaim ), "claimInfoEntry" , claimInfo )
203+ logger .V (6 ).Info ("Found existing claim info cache entry" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name , "claim" , klog .KObj (resourceClaim ), "claimInfoEntry" , claimInfo )
204204 }
205205
206206 // Add a reference to the current pod in the claim info.
@@ -216,7 +216,7 @@ func (m *ManagerImpl) PrepareResources(pod *v1.Pod) error {
216216
217217 // If this claim is already prepared, there is no need to prepare it again.
218218 if claimInfo .isPrepared () {
219- klog .V (5 ).InfoS ("Resources already prepared" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name , "claim" , klog .KObj (resourceClaim ))
219+ logger .V (5 ).Info ("Resources already prepared" , "pod" , klog .KObj (pod ), "podClaim" , podClaim .Name , "claim" , klog .KObj (resourceClaim ))
220220 return nil
221221 }
222222
@@ -250,7 +250,7 @@ func (m *ManagerImpl) PrepareResources(pod *v1.Pod) error {
250250 if err != nil {
251251 return fmt .Errorf ("failed to get gRPC client for driver %s: %w" , driverName , err )
252252 }
253- response , err := client .NodePrepareResources (context . Background () , & drapb.NodePrepareResourcesRequest {Claims : claims })
253+ response , err := client .NodePrepareResources (ctx , & drapb.NodePrepareResourcesRequest {Claims : claims })
254254 if err != nil {
255255 // General error unrelated to any particular claim.
256256 return fmt .Errorf ("NodePrepareResources failed: %w" , err )
@@ -338,7 +338,6 @@ func (m *ManagerImpl) GetResources(pod *v1.Pod, container *v1.Container) (*Conta
338338 // was generated for the referenced claim. There are valid use
339339 // cases when this might happen, so we simply skip it.
340340 if claimName == nil {
341- klog .V (5 ).InfoS ("No CDI devices, no claim generated" , "pod" , klog .KObj (pod ), "podClaimName" , podClaim .Name )
342341 continue
343342 }
344343 for _ , claim := range container .Resources .Claims {
@@ -362,16 +361,14 @@ func (m *ManagerImpl) GetResources(pod *v1.Pod, container *v1.Container) (*Conta
362361 }
363362 }
364363 }
365-
366- klog .V (5 ).InfoS ("Determined CDI devices for pod" , "pod" , klog .KObj (pod ), "cdiDevices" , cdiDevices )
367364 return & ContainerInfo {CDIDevices : cdiDevices }, nil
368365}
369366
370367// UnprepareResources calls a driver's NodeUnprepareResource API for each resource claim owned by a pod.
371368// This function is idempotent and may be called multiple times against the same pod.
372369// As such, calls to the underlying NodeUnprepareResource API are skipped for claims that have
373370// already been successfully unprepared.
374- func (m * ManagerImpl ) UnprepareResources (pod * v1.Pod ) error {
371+ func (m * ManagerImpl ) UnprepareResources (ctx context. Context , pod * v1.Pod ) error {
375372 var claimNames []string
376373 for i := range pod .Spec .ResourceClaims {
377374 claimName , _ , err := resourceclaim .Name (pod , & pod .Spec .ResourceClaims [i ])
@@ -386,10 +383,11 @@ func (m *ManagerImpl) UnprepareResources(pod *v1.Pod) error {
386383 }
387384 claimNames = append (claimNames , * claimName )
388385 }
389- return m .unprepareResources (pod .UID , pod .Namespace , claimNames )
386+ return m .unprepareResources (ctx , pod .UID , pod .Namespace , claimNames )
390387}
391388
392- func (m * ManagerImpl ) unprepareResources (podUID types.UID , namespace string , claimNames []string ) error {
389+ func (m * ManagerImpl ) unprepareResources (ctx context.Context , podUID types.UID , namespace string , claimNames []string ) error {
390+ logger := klog .FromContext (ctx )
393391 batches := make (map [string ][]* drapb.Claim )
394392 claimNamesMap := make (map [types.UID ]string )
395393 for _ , claimName := range claimNames {
@@ -445,7 +443,7 @@ func (m *ManagerImpl) unprepareResources(podUID types.UID, namespace string, cla
445443 if err != nil {
446444 return fmt .Errorf ("get gRPC client for DRA driver %s: %w" , driverName , err )
447445 }
448- response , err := client .NodeUnprepareResources (context . Background () , & drapb.NodeUnprepareResourcesRequest {Claims : claims })
446+ response , err := client .NodeUnprepareResources (ctx , & drapb.NodeUnprepareResourcesRequest {Claims : claims })
449447 if err != nil {
450448 // General error unrelated to any particular claim.
451449 return fmt .Errorf ("NodeUnprepareResources failed: %w" , err )
@@ -473,7 +471,7 @@ func (m *ManagerImpl) unprepareResources(podUID types.UID, namespace string, cla
473471 for _ , claimName := range claimNamesMap {
474472 claimInfo , _ := m .cache .get (claimName , namespace )
475473 m .cache .delete (claimName , namespace )
476- klog .V (6 ).InfoS ("Deleted claim info cache entry" , "claim" , klog .KRef (namespace , claimName ), "claimInfoEntry" , claimInfo )
474+ logger .V (6 ).Info ("Deleted claim info cache entry" , "claim" , klog .KRef (namespace , claimName ), "claimInfoEntry" , claimInfo )
477475 }
478476
479477 // Atomically sync the cache back to the checkpoint.
0 commit comments