HBASE-30095 Remove duplicate MiniDFSCluster startup in TestZooKeeper

[Sigma-Ma]

JIRA: https://issues.apache.org/jira/browse/HBASE-30095

What changes were proposed in this pull request?

Remove the duplicate TEST_UTIL.startMiniDFSCluster(2) call from TestZooKeeper#setUpBeforeClass.

The first MiniDFSCluster startup is retained, preserving the intended test initialization sequence: start MiniDFSCluster, start MiniZooKeeper, configure the test environment, and then start MiniHBaseCluster for each test.

Why are the changes needed?

HBaseTestingUtil should manage only one MiniDFSCluster at a time. Calling startMiniDFSCluster twice replaces the stored dfsCluster and dfsClusterFixer references without shutting down the first cluster.

This can leave the first cluster running against the same test directories and cause HDFS namespace conflicts, including InconsistentFSStateException, in CI runs.

How was this patch tested?

mvn -pl hbase-server -am -Dtest=TestZooKeeper -Dsurefire.rerunFailingTestsCount=0 test

[Umeshkumar9414]

This was intentionally added by @saintstack here 7df9490. Although I didn't get it, he can help us here.

[Sigma-Ma]

This was intentionally added by @saintstack here 7df9490. Although I didn't get it, he can help us here.

Thanks for pointing this out. I checked the original commit, and its message says the TestZooKeeper change was “probably useless” and that the actual issue was related to the ZooKeeper startup check hanging.

The test already starts a MiniDFSCluster earlier in setUpBeforeClass, so the added line starts it twice on the same TEST_UTIL. That’s why this PR removes only the second call.

@saintstack, please let me know if I’m missing any context here.

[Sigma-Ma]

Hi @haridsv, would you mind taking a look when you have time, or suggesting an appropriate reviewer? This is one of the remaining fixes split from HBASE-30073. Thanks!