@@ -64,7 +64,7 @@ func NewComputeDomainManager(config *ManagerConfig) *ComputeDomainManager {
6464 factory := nvinformers .NewSharedInformerFactory (config .clientsets .Nvidia , informerResyncPeriod )
6565 informer := factory .Resource ().V1beta1 ().ComputeDomains ().Informer ()
6666
67- klog .Infof ("Creating new ComputeDomainManager with config %+v " , config )
67+ klog .Infof ("Creating new ComputeDomainManager for %s/%s " , config . driverName , config . driverNamespace )
6868 m := & ComputeDomainManager {
6969 config : config ,
7070 factory : factory ,
@@ -149,7 +149,7 @@ func (m *ComputeDomainManager) Get(uid string) (*nvapi.ComputeDomain, error) {
149149 return nil , fmt .Errorf ("error retrieving ComputeDomain by UID: %w" , err )
150150 }
151151 if len (cds ) == 0 {
152- klog .Infof ("No ComputeDomain found with UID: %s" , uid )
152+ klog .V ( 2 ). Infof ("No ComputeDomain found with UID: %s" , uid )
153153 return nil , nil
154154 }
155155 if len (cds ) != 1 {
@@ -169,7 +169,7 @@ func (m *ComputeDomainManager) RemoveFinalizer(ctx context.Context, uid string)
169169 return fmt .Errorf ("error retrieving ComputeDomain: %w" , err )
170170 }
171171 if cd == nil {
172- klog .Infof ("ComputeDomain with UID %s not found, nothing to do" , uid )
172+ klog .V ( 2 ). Infof ("ComputeDomain with UID %s not found, nothing to do" , uid )
173173 return nil
174174 }
175175
@@ -185,16 +185,31 @@ func (m *ComputeDomainManager) RemoveFinalizer(ctx context.Context, uid string)
185185 }
186186 }
187187 if len (cd .Finalizers ) == len (newCD .Finalizers ) {
188- klog .Infof ("Finalizer not found on ComputeDomain %s/%s, nothing to do" , cd .Namespace , cd .Name )
188+ klog .V ( 2 ). Infof ("Finalizer not found on ComputeDomain %s/%s, nothing to do" , cd .Namespace , cd .Name )
189189 return nil
190190 }
191191
192192 if _ , err = m .config .clientsets .Nvidia .ResourceV1beta1 ().ComputeDomains (cd .Namespace ).Update (ctx , newCD , metav1.UpdateOptions {}); err != nil {
193193 return fmt .Errorf ("error updating ComputeDomain: %w" , err )
194194 }
195+
195196 return nil
196197}
197198
199+ // logNodesWithComputeDomainLabel logs nodes that have a ComputeDomain label and returns their names
200+ func (m * ComputeDomainManager ) logNodesWithComputeDomainLabel (nodes * corev1.NodeList , cdUID string ) []string {
201+ if len (nodes .Items ) == 0 {
202+ klog .Infof ("No nodes found with label for ComputeDomain with UID %s" , cdUID )
203+ return nil
204+ }
205+
206+ nodeNames := []string {}
207+ for _ , node := range nodes .Items {
208+ nodeNames = append (nodeNames , node .Name )
209+ }
210+ return nodeNames
211+ }
212+
198213// AssertWorkloadsCompletes ensures that all workloads asssociated with a ComputeDomain have completed.
199214//
200215// TODO: We should probably also check to ensure that all ResourceClaims
@@ -219,38 +234,9 @@ func (m *ComputeDomainManager) AssertWorkloadsCompleted(ctx context.Context, cdU
219234 }
220235
221236 if len (nodes .Items ) != 0 {
222- // show nodes with labels
223- nodeNames := []string {}
224- for _ , node := range nodes .Items {
225- nodeNames = append (nodeNames , node .Name )
226- }
227- klog .Errorf ("Found %d nodes with label for ComputeDomain with UID %s: %v" , len (nodes .Items ), cdUID , nodeNames )
228- return fmt .Errorf ("nodes exist with label for ComputeDomain %s" , cdUID )
237+ nodeNames := logNodesWithComputeDomainLabel (nodes , cdUID )
238+ return fmt .Errorf ("nodes %v with label for ComputeDomain %s" , nodeNames , cdUID )
229239 }
230-
231- // check if all resource claims for workloads are gone
232- cd , err := m .Get (cdUID )
233- if err != nil {
234- return fmt .Errorf ("error retrieving ComputeDomain: %w" , err )
235- }
236-
237- resourceClaims , err := m .config .clientsets .Core .ResourceV1beta1 ().ResourceClaims (cd .Namespace ).List (ctx , metav1.ListOptions {
238- LabelSelector : metav1 .FormatLabelSelector (labelSelector ),
239- })
240- if err != nil {
241- return fmt .Errorf ("error retrieving ResourceClaims: %w" , err )
242- }
243-
244- if len (resourceClaims .Items ) != 0 {
245- claimNames := []string {}
246- for _ , claim := range resourceClaims .Items {
247- claimNames = append (claimNames , claim .Name )
248- }
249- klog .Errorf ("Found %d ResourceClaims for ComputeDomain with UID %s: %v" ,
250- len (resourceClaims .Items ), cdUID , claimNames )
251- return fmt .Errorf ("ResourceClaims exist for ComputeDomain %s" , cdUID )
252- }
253-
254240 return nil
255241}
256242
0 commit comments