[fix][admin] Set local policies overwrites "number of bundles" passed during namespace creation

[oneby-wang]

Fixes #23897

Motivation

Set local policies will overwrite "number of bundles" passed during namespace creation. If we don't call any get namespace policies operation, local policies will not be created, so it will be overwritten.

pulsar/pulsar-broker/src/main/java/org/apache/pulsar/common/naming/NamespaceBundleFactory.java

Lines 164 to 182 in 8879b3b

	private CompletableFuture<NamespaceBundles> copyToLocalPolicies(NamespaceName namespace) {
	return pulsar.getPulsarResources().getNamespaceResources().getPoliciesAsync(namespace)
	.thenCompose(optPolicies -> {
	if (!optPolicies.isPresent()) {
	return CompletableFuture.completedFuture(getBundles(namespace, Optional.empty()));
	}
	Policies policies = optPolicies.get();
	LocalPolicies localPolicies = new LocalPolicies(policies.bundles,
	null,
	null);
	return pulsar.getPulsarResources().getLocalPolicies()
	.createLocalPoliciesAsync(namespace, localPolicies)
	.thenApply(stat -> getBundles(namespace,
	Optional.of(Pair.of(localPolicies, 0L))));
	});
	}

The following steps can reproduce the problem.

1. pulsar-admin namespaces create t/n --bundles 10
2. pulsar-admin namespaces set-bookie-affinity-group --primary-group temp t/n
3. pulsar-admin namespaces bundles t/n

But the following steps will not reproduce the problem.

1. pulsar-admin namespaces create t/n --bundles 10
2. pulsar-admin namespaces bundles t/n
3. pulsar-admin namespaces set-bookie-affinity-group --primary-group temp t/n
4. pulsar-admin namespaces bundles t/n

Modifications

If local policies is not present, first try to use namespace policies bundle data, then fallback to config().getDefaultNumberOfNamespaceBundles()

Verifying this change

• Make sure that the change passes the CI checks.

Does this pull request potentially affect one of the following parts:

If the box was checked, please highlight the changes

• Dependencies (add or upgrade a dependency)
• The public API
• The schema
• The default values of configurations
• The threading model
• The binary protocol
• The REST endpoints
• The admin CLI options
• The metrics
• Anything that affects deployment

Documentation

• doc
• doc-required
• doc-not-needed
• doc-complete

Matching PR in forked repository

PR in forked repository: oneby-wang#5

[github-actions]

@oneby-wang Please add the following content to your PR description and select a checkbox:

- [ ] `doc` <!-- Your PR contains doc changes -->
- [ ] `doc-required` <!-- Your PR changes impact docs and you will update later -->
- [ ] `doc-not-needed` <!-- Your PR changes do not impact docs -->
- [ ] `doc-complete` <!-- Docs have been already added -->

[BewareMyPower]

From the PR description, it seems possible to add a test?

[oneby-wang]

From the PR description, it seems possible to add a test?

Thanks for advice, I'll add unit tests.

[lhotari]

LGTM

[oneby-wang]

ExtensibleLoadManagerTest.testAntiaffinityPolicy() unit test failure in fork repository PR.
https://github.com/oneby-wang/pulsar/actions/runs/17970886848/job/51114222121?pr=5

Seems this default bundle config affects this unit test. Previous: static default 1 numBundles. After: namespace policies numBundles.

pulsar/pulsar-broker/src/main/java/org/apache/pulsar/broker/admin/impl/NamespacesBase.java

Lines 1792 to 1793 in 7c8d3c9

	.orElseGet(() -> new LocalPolicies(defaultBundle(),
	null, antiAffinityGroup))

PoliciesUtil.defaultBundle() method.

pulsar/pulsar-common/src/main/java/org/apache/pulsar/common/policies/data/PoliciesUtil.java

Lines 33 to 41 in 7c8d3c9

	public static BundlesData defaultBundle() {
	List<String> boundaries = new ArrayList<>();
	boundaries.add(FIRST_BOUNDARY);
	boundaries.add(LAST_BOUNDARY);
	return BundlesData.builder()
	.numBundles(1)
	.boundaries(boundaries)
	.build();
	}

Seems some bugs in ExtensibleLoadManager or ExtensibleLoadManagerTest?
https://pulsar.apache.org/docs/3.0.x/administration-load-balance/#distribute-anti-affinity-namespaces-across-failure-domains

Not familiar with ExtensibleLoadManager, may give me some entry point, may be I can help fix this problem? Seems here?

pulsar/pulsar-broker/src/main/java/org/apache/pulsar/broker/namespace/NamespaceService.java

Lines 224 to 237 in 7c8d3c9

	public CompletableFuture<Optional<LookupResult>> getBrokerServiceUrlAsync(TopicName topic, LookupOptions options) {
	long startTime = System.nanoTime();
	CompletableFuture<Optional<LookupResult>> future = getBundleAsync(topic)
	.thenCompose(bundle -> {
	// Do redirection if the cluster is in rollback or deploying.
	return findRedirectLookupResultAsync(bundle).thenCompose(optResult -> {
	if (optResult.isPresent()) {
	LOG.info("[{}] Redirect lookup request to {} for topic {}",
	pulsar.getBrokerId(), optResult.get(), topic);
	return CompletableFuture.completedFuture(optResult);
	}
	if (ExtensibleLoadManagerImpl.isLoadManagerExtensionEnabled(pulsar)) {
	return loadManager.get().findBrokerServiceUrl(Optional.of(topic), bundle, options);

[oneby-wang]

Seems the anti-affinity works at the bundle level.

pulsar/pulsar-broker/src/main/java/org/apache/pulsar/broker/loadbalance/impl/LoadManagerShared.java

Lines 588 to 620 in 7c8d3c9

	public static CompletableFuture<Map<String, Integer>> getAntiAffinityNamespaceOwnedBrokers(
	final PulsarService pulsar, final String namespaceName,
	Set<Map.Entry<String, ServiceUnitStateData>> bundleOwnershipData) {
	CompletableFuture<Map<String, Integer>> antiAffinityNsBrokersResult = new CompletableFuture<>();
	getNamespaceAntiAffinityGroupAsync(pulsar, namespaceName)
	.thenAccept(antiAffinityGroupOptional -> {
	if (antiAffinityGroupOptional.isEmpty()) {
	antiAffinityNsBrokersResult.complete(null);
	return;
	}
	final String antiAffinityGroup = antiAffinityGroupOptional.get();
	final Map<String, Integer> brokerToAntiAffinityNamespaceCount = new ConcurrentHashMap<>();
	final List<CompletableFuture<Void>> futures = new ArrayList<>();
	bundleOwnershipData
	.forEach(etr -> {
	var stateData = etr.getValue();
	var bundle = etr.getKey();
	if (stateData.state() == ServiceUnitState.Owned
	&& StringUtils.isNotBlank(stateData.dstBroker())) {
	CompletableFuture<Void> future = new CompletableFuture<>();
	futures.add(future);
	countAntiAffinityNamespaceOwnedBrokers
	(stateData.dstBroker(),
	LoadManagerShared.getNamespaceNameFromBundleName(bundle),
	future, pulsar,
	antiAffinityGroup, brokerToAntiAffinityNamespaceCount);
	}
	});
	FutureUtil.waitForAll(futures)
	.thenAccept(r -> antiAffinityNsBrokersResult.complete(brokerToAntiAffinityNamespaceCount));

I think this unit test may work in a wrong way due to this PR related issue, so I changed numBundles to 1 in createNamespace method.

pulsar/tests/integration/src/test/java/org/apache/pulsar/tests/integration/loadbalance/ExtensibleLoadManagerTest.java

Lines 315 to 321 in 7c8d3c9

	for (int i = 0; i < activeBrokers.size(); i++) {
	String namespace = antiAffinityEnabledNameSpace + "-" + i;
	admin.namespaces().createNamespace(namespace, 10);
	admin.namespaces().setNamespaceAntiAffinityGroup(namespace, namespaceAntiAffinityGroup);
	admin.clusters().createFailureDomain(clusterName, namespaceAntiAffinityGroup, FailureDomain.builder()
	.brokers(Set.of(activeBrokers.get(i))).build());
	}