Conversation
…ntroller The StackitCluster reconciler now owns the whole apiserver target pool and derives it from the control-plane Machines of the cluster.
A cluster deletion filters out every machine at once, and the placeholder must not overwrite the pool while the API server is still needed.
tuunit
changed the title
tuunit
requested changes
Sep 25, 2026
Comment on lines
+31
to
+32
| // maxTargetNameLength is the STACKIT limit on a target display name. | ||
| maxTargetNameLength = 63 |
Collaborator
There was a problem hiding this comment.
thats the same length as kubernetes resources anyway
https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#dns-label-names
Comment on lines
+96
to
+127
| // targetName turns a machine name into a valid STACKIT target display name: | ||
| // letters, digits and inner hyphens, at most 63 characters. A qualifying name is | ||
| // returned unchanged; any other carries a digest so that two machines cannot | ||
| // collapse onto one target. | ||
| func targetName(machineName string) string { | ||
| var sanitized strings.Builder | ||
| previousHyphen := false | ||
| for _, r := range machineName { | ||
| switch { | ||
| case (r >= '0' && r <= '9') || (r >= 'a' && r <= 'z') || (r >= 'A' && r <= 'Z'): | ||
| sanitized.WriteRune(r) | ||
| previousHyphen = false | ||
| case !previousHyphen: | ||
| sanitized.WriteByte('-') | ||
| previousHyphen = true | ||
| } | ||
| } | ||
| name := strings.Trim(sanitized.String(), "-") | ||
| if name == machineName && len(name) <= maxTargetNameLength { | ||
| return name | ||
| } | ||
|
|
||
| sum := sha256.Sum256([]byte(machineName)) | ||
| digest := hex.EncodeToString(sum[:])[:targetNameDigestLength] | ||
| if name == "" { | ||
| return digest | ||
| } | ||
| suffix := "-" + digest | ||
| if len(name) > maxTargetNameLength-len(suffix) { | ||
| name = strings.TrimRight(name[:maxTargetNameLength-len(suffix)], "-") | ||
| } | ||
| return name + suffix |
Collaborator
There was a problem hiding this comment.
as kubernetes machine names are forced to follow RFC1123 as mentioned above. capital letters or whitespace isn't possible anyways. so why do we need this sanitization step?
Comment on lines
+610
to
+611
| // Control plane status updates are frequent, and each one reaches this path. | ||
| if targetPool.GetTargetPort() == port && sameTargets(targetPool.GetTargets(), desired) { |
Collaborator
There was a problem hiding this comment.
what is this comment supposed to tell me? so yes this happens frequently but why do i need to know this?
| // STACKIT NLB target pools must contain at least one target. Leave the | ||
| // last target in place; deleting the load balancer removes it. | ||
| return nil | ||
| keys = append(keys, target.GetDisplayName()+"\x00"+target.GetIp()) |
Collaborator
There was a problem hiding this comment.
So whats the "magic" nullbyte for? This requires a comment or documentation link or something
Comment on lines
196
to
+201
| Watches(&clusterv1.Cluster{}, handler.EnqueueRequestsFromMapFunc(r.stackitClusterRequestsForCluster)). | ||
| Watches( | ||
| &clusterv1.Machine{}, | ||
| handler.EnqueueRequestsFromMapFunc(r.stackitClusterRequestsForMachine), | ||
| builder.WithPredicates(predicate.Funcs{UpdateFunc: machineTargetPoolChanged}), | ||
| ). |
Collaborator
There was a problem hiding this comment.
This naming scheme seems a bit off lets take the opportunity to get something less verbose in place:
Suggested change
| Watches(&clusterv1.Cluster{}, handler.EnqueueRequestsFromMapFunc(r.stackitClusterRequestsForCluster)). | |
| Watches( | |
| &clusterv1.Machine{}, | |
| handler.EnqueueRequestsFromMapFunc(r.stackitClusterRequestsForMachine), | |
| builder.WithPredicates(predicate.Funcs{UpdateFunc: machineTargetPoolChanged}), | |
| ). | |
| Watches(&clusterv1.Cluster{}, handler.EnqueueRequestsFromMapFunc(r.FilterForStackitClusterResources)). | |
| Watches( | |
| &clusterv1.Machine{}, | |
| handler.EnqueueRequestsFromMapFunc(r.LoadBalancerTargetsForControlPlanes), | |
| builder.WithPredicates(predicate.Funcs{UpdateFunc: machineTargetPoolChanged}), | |
| ). |
As the comment of stackitClusterRequestsForMachine already implies its only concerned about the API servers load balancer for the control planes so we should indicate this in the name. I would suggest:
LoadBalancerTargetsForControlPlanes
And while we are at it: stackitClusterRequestsForCluster -> FilterForStackitClusterResources
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #24
The API server load balancer is created by the
StackitClusterreconciler while its targetpool was maintained by
StackitMachine. That leaked cluster-level infrastructure into themachine controller and let several machines write to the same pool concurrently. STACKIT takes
an explicit target list rather than label-based membership, so the pool now has exactly one
owner.
StackitClusterwatches control-planeMachineevents and rebuilds the pool fromMachine.Status.Addresses, which Cluster API propagates fromStackitMachine.status.addresses.No RBAC change. The watch carries a predicate, since control plane status updates are frequent
and each one would otherwise cost a full reconcile.
EnsureAPIServerLoadBalancerTargetandDeleteAPIServerLoadBalancerTargetare replaced bySetAPIServerLoadBalancerTargets, which applies the pool as a set and skips the write whennothing changed.
reconcileAPIServerLoadBalancerTarget,deleteAPIServerLoadBalancerTargetand
EnsureForMachineare gone.retryable. Machine names are therefore sanitized into valid display names, with a digest
against collisions, and duplicate target IPs are dropped.
plane before the infrastructure, so every machine is filtered out at once while the API server
is still needed for drain and etcd member removal.
sets only
LoadBalancerReady, neverReady. The machine reconciler stops while the cluster isnot ready, so failing the cluster would block replacing the machine whose target is broken.
Accepted trade-off: removing a target is no longer a precondition of the machine finalizer.
If the cluster controller is wedged, a target pointing at a dead IP survives, and the create path
sets no
ActiveHealthCheckthat would fail it out. Should STACKIT recycle that IP, the loadbalancer forwards API traffic to an unrelated VM.
make lint,make testandmake manifestsare clean,config/is unchanged. The specs for thepool collapse and the permanent-failure case were checked against the unfixed code and fail there.
E2E was not run.