[fix][admin] Reject empty cluster migration URL: fix inverted ClusterUrl.isEmpty()

[SongOf]

Motivation

ClusterPolicies.ClusterUrl#isEmpty() has its condition inverted:

return serviceUrl != null && serviceUrlTls != null && brokerServiceUrl == null && brokerServiceUrlTls == null;

It only returns true when both HTTP URLs are set and both broker URLs are unset, and returns false for a ClusterUrl whose four fields are all null.

Its only caller is the cluster-migration admin endpoint (ClustersBase#updateClusterMigration), which uses it to reject migrated=true without a target:

if (isMigrated && clusterUrl.isEmpty()) {
    asyncResponse.resume(new RestException(Status.BAD_REQUEST, "Cluster url must not be empty"));

Because of the inverted predicate this guard never fires for the case it exists for. pulsar-admin clusters update-cluster-migration --migrated with no --service-url/--broker-url options is accepted and persisted; later, when the broker redirects producers/consumers with CommandTopicMigrated, the client receives null URLs and fails in HandlerState#setRedirectedClusterURI (logged, not surfaced), so the affected producers/consumers reconnect in a loop with no clear error. Conversely, a legitimate ClusterUrl carrying only the HTTP endpoints is rejected as "empty".

Modifications

• ClusterPolicies.ClusterUrl#isEmpty() now returns true only when none of the four URLs is set (null or blank). Any single configured URL — HTTP or binary, plain or TLS — makes it non-empty. No signature, wire, or REST change; the guard in ClustersBase#updateClusterMigration is unchanged and now behaves as its error message says.

Verifying this change

This change added tests and can be verified as follows:

• ClusterUrlTest (pulsar-client-admin-api) — pins the predicate: all-null / all-blank is empty; HTTP-only, broker-only, and fully populated URLs are not. Fails on the unpatched code (testNoUrlIsEmpty, testBlankUrlsAreEmpty, testHttpOnlyUrlIsNotEmpty).
• AdminTest#clusters (pulsar-broker) — adds the end-to-end assertion at the admin resource: updateClusterMigration(..., migrated=true, new ClusterUrl()) returns 400 BAD_REQUEST, and a ClusterUrl with only a broker service URL is accepted and read back via getClusterMigration. On the unpatched code the empty-URL call is accepted and the test fails at fail("should have failed"). The validateSuperUserAccessAsync invocation count in this test is bumped 24 → 26 for the two accepted calls (the rejected call fails at the guard before any authorization check).

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

This is an internal behavior change: the only user-visible effect is that POST /admin/v2/clusters/{cluster}/migrate?migrated=true with an empty body/URL now returns 400 as documented, instead of being accepted. No public API, schema, configuration, wire protocol, REST endpoint shape, CLI option, or metric changes.

Documentation

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