KAFKA-18000: Fix flaky ReplicaManagerTest#testSuccessfulBuildRemoteLo…

…gAuxStateMetrics (#17887)


Reviewers: David Arthur <[email protected]>

core/src/test/scala/unit/kafka/server/ReplicaManagerTest.scala

@@ -51,7 +51,6 @@ import org.apache.kafka.common.requests.FetchRequest.PartitionData
 import org.apache.kafka.common.requests.ProduceResponse.PartitionResponse
 import org.apache.kafka.common.requests._
 import org.apache.kafka.common.security.auth.KafkaPrincipal
-import org.apache.kafka.common.test.api.Flaky
 import org.apache.kafka.common.utils.{Exit, LogContext, Time, Utils}
 import org.apache.kafka.coordinator.transaction.TransactionLogConfig
 import org.apache.kafka.image._
@@ -4343,7 +4342,6 @@ class ReplicaManagerTest {
     }
   }
 
-  @Flaky("KAFKA-18000")
   @Test
   def testSuccessfulBuildRemoteLogAuxStateMetrics(): Unit = {
     val tp0 = new TopicPartition(topic, 0)
@@ -4390,10 +4388,12 @@ class ReplicaManagerTest {
       replicaManager.becomeLeaderOrFollower(0, leaderAndIsrRequest, (_, _) => ())
 
       // Replicas fetch from the leader periodically, therefore we check that the metric value is increasing
-      assertTrue(brokerTopicStats.topicStats(tp0.topic()).buildRemoteLogAuxStateRequestRate.count > 0)
+      waitUntilTrue(() => brokerTopicStats.topicStats(tp0.topic()).buildRemoteLogAuxStateRequestRate.count > 0,
+        "Should have buildRemoteLogAuxStateRequestRate count > 0, but got:" + brokerTopicStats.topicStats(tp0.topic()).buildRemoteLogAuxStateRequestRate.count)
       assertEquals(0, brokerTopicStats.topicStats(tp0.topic()).failedBuildRemoteLogAuxStateRate.count)
       // Verify aggregate metrics
-      assertTrue(brokerTopicStats.allTopicsStats.buildRemoteLogAuxStateRequestRate.count > 0)
+      waitUntilTrue(() => brokerTopicStats.allTopicsStats.buildRemoteLogAuxStateRequestRate.count > 0,
+        "Should have all topic buildRemoteLogAuxStateRequestRate count > 0, but got:" + brokerTopicStats.allTopicsStats.buildRemoteLogAuxStateRequestRate.count)
       assertEquals(0, brokerTopicStats.allTopicsStats.failedBuildRemoteLogAuxStateRate.count)
     } finally {
       replicaManager.shutdown(checkpointHW = false)