Skip to content

Replacing endpointSlice Reconciler with a direct Pod Reconciler #300

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 16 commits into from
Feb 13, 2025
Merged

Replacing endpointSlice Reconciler with a direct Pod Reconciler #300

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
2133118
reversion to pod reconciliation
kfswain Jan 31, 2025
beac309
adding ready check and unit tests
kfswain Jan 31, 2025
31de8a9
updating test
kfswain Jan 31, 2025
5e21da1
ablating unnecessary func
kfswain Jan 31, 2025
6ac1e10
embedding ready status into update so non-ready pods are deleted
kfswain Jan 31, 2025
f9e0807
scrubbing serviceName & zone as they are obsolete
kfswain Feb 11, 2025
e989f7f
implementing pod cache flushing logic
kfswain Feb 12, 2025
72c2d63
Renaming file so merge confilcts can find the diffs easier
kfswain Feb 12, 2025
62dcdb6
cleaning up messy merge conflict
kfswain Feb 12, 2025
9f61997
nil checking short circuit
kfswain Feb 13, 2025
e6797ba
Listing fixes
kfswain Feb 13, 2025
83ed549
feedback cleanup
kfswain Feb 13, 2025
adf9f8e
log formatting and removing pods if not found
kfswain Feb 13, 2025
48693e8
removing err to provent perma-reconciliation
kfswain Feb 13, 2025
7c341eb
removing dev image ref
kfswain Feb 13, 2025
b244225
cleaning up err logic
kfswain Feb 13, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 2 additions & 5 deletions pkg/ext-proc/backend/datastore.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,19 +130,16 @@ func (ds *K8sDatastore) flushPodsAndRefetch(ctx context.Context, ctrlClient clie
}); err != nil {
klog.Error(err, "error listing clients")
}
podMap := sync.Map{}
ds.pods.Clear()

for _, k8sPod := range podList.Items {
pod := Pod{
Name: k8sPod.Name,
Address: k8sPod.Status.PodIP + ":" + strconv.Itoa(int(newServerPool.Spec.TargetPortNumber)),
}
podMap.Store(pod, true)
ds.pods.Store(pod, true)
}

// Clear is thread-safe, so if there are any in-flight accesses this should prevent odd data behavior. Then we replace after.
ds.pods.Clear()
ds.pods = &podMap
}

func selectorFromInferencePoolSelector(selector map[v1alpha1.LabelKey]v1alpha1.LabelValue) labels.Selector {
Expand Down
4 changes: 3 additions & 1 deletion pkg/ext-proc/backend/inferencepool_reconciler.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,9 +36,11 @@ func (c *InferencePoolReconciler) Reconcile(ctx context.Context, req ctrl.Reques
return ctrl.Result{}, err
}
if c.Datastore.inferencePool == nil || !reflect.DeepEqual(serverPool.Spec.Selector, c.Datastore.inferencePool.Spec.Selector) {
c.updateDatastore(serverPool)
c.Datastore.flushPodsAndRefetch(ctx, c.Client, serverPool)
} else {
c.updateDatastore(serverPool)
}
c.updateDatastore(serverPool)

return ctrl.Result{}, nil
}
Expand Down
8 changes: 5 additions & 3 deletions pkg/ext-proc/backend/pod_reconciler.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,6 @@ package backend
import (
"context"
"strconv"
"time"

corev1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/runtime"
Expand All @@ -26,7 +25,10 @@ func (c *PodReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.R
inferencePool, err := c.Datastore.getInferencePool()
if err != nil {
klog.V(logutil.DEFAULT).Infof("Skipping reconciling Pod because the InferencePool is not available yet: %v", err)
return ctrl.Result{Requeue: true, RequeueAfter: time.Second}, nil
// When the inferencePool is initialized it lists the appropriate pods and populates the datastore, so no need to requeue.
return ctrl.Result{}, nil
} else if inferencePool.Namespace != req.Namespace {
return ctrl.Result{}, nil
}

klog.V(logutil.VERBOSE).Info("reconciling Pod", req.NamespacedName)
Expand All @@ -53,7 +55,7 @@ func (c *PodReconciler) updateDatastore(k8sPod *corev1.Pod, inferencePool *v1alp
Name: k8sPod.Name,
Address: k8sPod.Status.PodIP + ":" + strconv.Itoa(int(inferencePool.Spec.TargetPortNumber)),
}
if !c.Datastore.LabelsMatch(k8sPod.ObjectMeta.Labels) || !podIsReady(k8sPod) {
if !k8sPod.DeletionTimestamp.IsZero() || !c.Datastore.LabelsMatch(k8sPod.ObjectMeta.Labels) || !podIsReady(k8sPod) {
c.Datastore.pods.Delete(pod)
} else {
c.Datastore.pods.Store(pod, true)
Expand Down