[d-wang-commit]

Description

Count the full DataRegion leader population

The leader-distribution fixture creates three or six user DataRegionGroups. Internal database groups participate in the same balancing round, but the assertions removed those groups and then required the remaining user-only subset to be balanced. A balanced cluster can therefore fail with a user-only distribution such as {1=1, 3=2}.

Count the fixture's DataRegionGroups including internal databases, without mixing in independently balanced SchemaRegion leaders. Poll for exactly one Running leader per group, the expected number of leader-holding nodes, and a maximum per-node count difference of one. Keep the ReadOnly and stopped-node transitions and verify the same conditions after them.

Check only the migrated consensus group

A migration removes one RegionGroup replica, so other groups' consensus directories may remain on the source DataNode. Replace the whole-directory emptiness check and hardcoded 1_1 exception with checks for the selected group ID. Resolve Ratis group directories through the existing Raft group UUID conversion. Add a regression that leaves an unrelated three-replica database on the source node and verifies that its directories survive the migration.

Tests

• UT: none added; this PR only changes integration-test fixtures and assertions.
• IT: updated IoTDBRegionGroupLeaderDistributionIT and the shared migration assertions; added migrationShouldPreserveOtherRegionFiles in IoTDBRegionMigrateNormalForIoTV2StreamIT.
• Passed all three methods in IoTDBRegionGroupLeaderDistributionIT and all three methods in IoTDBRegionMigrateNormalForIoTV2StreamIT with a clean reactor build: mvn clean verify -pl integration-test -am -PClusterIT,with-integration-tests -DskipUTs -Dit.test=IoTDBRegionGroupLeaderDistributionIT,IoTDBRegionMigrateNormalForIoTV2StreamIT -DfailIfNoTests=false -Dfailsafe.failIfNoSpecifiedTests=false.
• Also passed IoTDBRegionMigrateNormalITForIoTV2BatchIT#normal1C2DTest and IoTDBRegionMigrateAddingPeerCrashForRatisIT#addingPeerCrashShouldFailAndRollback: 8 selected IT methods passed in total.
• The same six leader/stream methods passed in the downstream edition. Both builds used the existing Procedure persistence implementation; no production fix was included.
• Spotless and the clean English reactor build passed. No production or localized messages were changed.

Side effects and risks

No production code, persistence format, leader-balancing policy, or node-state restriction changes. The leader checks retain the existing 50-second bound and now reject missing leaders and non-Running leader replicas. Migration checks still require removal of the selected source directory and existence of the destination directory; unrelated RegionGroups no longer cause a false failure. A forced-down DataNode retains the existing skip behavior for local file checks.

---

This PR has:

• been self-reviewed.
• added comments explaining the assertion scope.
• added integration tests.
• been tested in a test IoTDB cluster.

---

Key changed/added classes (or packages if there are too many classes) in this PR

• IoTDBRegionGroupLeaderDistributionIT: include internal DataRegion leaders and reject incomplete or unhealthy distributions.
• IoTDBRegionOperationReliabilityITFramework: check directories for the selected RegionGroup.
• IoTDBRegionMigrateITFrameworkForRatis: map the selected group to its Ratis UUID directory.
• IoTDBRegionMigrateNormalForIoTV2StreamIT: verify unrelated region files survive migration.