github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2026-07-18 04:54:54 +00:00
Code Issues Projects Releases Wiki Activity

Fix scheduler flaky test: wait for DeviceClass cache sync in dynamicresources tests

Browse Source 
When DRAExtendedResource is enabled, the dynamicresources test setup
registers an event handler for DeviceClasses but was not waiting for it
to sync. This can lead to flaky tests where the cache is not fully
populated when the test starts.

This change captures the event handler registration and includes its
DoneChecker in a WaitFor call.
This commit is contained in:
Troy Chiu
2026-03-06 01:41:31 +00:00
parent 77c013637c
commit 1d2165b29c
1 changed files with 10 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
pkg/scheduler/framework/plugins/dynamicresources/dynamicresources_test.go
View File

@@ -2997,6 +2997,7 @@ func setup(tCtx ktesting.TContext, args *config.DynamicResourcesArgs, nodes []*v
tc.client.ReactionChain = append(apiReactors, tc.client.ReactionChain...)
tc.informerFactory = informers.NewSharedInformerFactory(tc.client, 0)
var doneCheckers []cache.DoneChecker
resourceSliceTrackerOpts := resourceslicetracker.Options{
EnableDeviceTaintRules: true,
SliceInformer: tc.informerFactory.Resource().V1().ResourceSlices(),
@@ -3006,6 +3007,7 @@ func setup(tCtx ktesting.TContext, args *config.DynamicResourcesArgs, nodes []*v
}
resourceSliceTracker, err := resourceslicetracker.StartTracker(tCtx, resourceSliceTrackerOpts)
require.NoError(tCtx, err, "couldn't start resource slice tracker")
doneCheckers = append(doneCheckers, resourceSliceTracker.HasSyncedChecker())
claimsCache := assumecache.NewAssumeCache(tCtx.Logger(), tc.informerFactory.Resource().V1().ResourceClaims().Informer(), "resource claim", "", nil)
// NewAssumeCache calls the informer's AddEventHandler method to register
@@ -3018,14 +3020,14 @@ func setup(tCtx ktesting.TContext, args *config.DynamicResourcesArgs, nodes []*v
// This is not the registered handler that is used by the DRA
// manager, but it is close enough because the assume cache
// uses a single boolean for "is synced" for all handlers.
registeredHandler := claimsCache.AddEventHandler(cache.ResourceEventHandlerFuncs{})
doneCheckers = append(doneCheckers, claimsCache.AddEventHandler(cache.ResourceEventHandlerFuncs{}).HasSyncedChecker())
tc.draManager = NewDRAManager(tCtx, claimsCache, resourceSliceTracker, tc.informerFactory)
if features.EnableDRAExtendedResource {
cache := tc.draManager.DeviceClassResolver().(*extendedresourcecache.ExtendedResourceCache)
if _, err := tc.informerFactory.Resource().V1().DeviceClasses().Informer().AddEventHandler(cache); err != nil {
tCtx.Logger().Error(err, "failed to add device class informer event handler")
}
deviceClassHandlerRegistration, err := tc.informerFactory.Resource().V1().DeviceClasses().Informer().AddEventHandler(cache)
require.NoError(tCtx, err, "failed to add device class informer event handler")
doneCheckers = append(doneCheckers, deviceClassHandlerRegistration.HasSyncedChecker())
}
opts := []runtime.Option{
@@ -3068,11 +3070,11 @@ func setup(tCtx ktesting.TContext, args *config.DynamicResourcesArgs, nodes []*v
})
tc.informerFactory.WaitForCacheSync(tCtx.Done())
// The above does not tell us if the registered handler (from NewAssumeCache)
// is synced, we need to wait until HasSynced of the handler returns
// true, this ensures that the assume cache is in sync with the informer's
// The above does not tell us if the registered handlers (e.g. from NewAssumeCache)
// are synced, we need to wait until the event handlers confirm that they are synced.
// This ensures that the assume cache is in sync with the informer's
// store which has been informed by at least one full LIST of the underlying storage.
cache.WaitForNamedCacheSyncWithContext(tCtx, registeredHandler.HasSynced, resourceSliceTracker.HasSynced)
cache.WaitFor(tCtx, "event handlers", doneCheckers...)
for _, node := range nodes {
nodeInfo := framework.NewNodeInfo()