soarez commented on a change in pull request #9577:
URL: https://github.com/apache/kafka/pull/9577#discussion_r760091660
##########
File path: core/src/main/scala/kafka/controller/KafkaController.scala
##########
@@ -2443,6 +2443,117 @@ class KafkaController(val config: KafkaConfig,
}
}
+ def alterReplicaState(alterReplicaStateRequest: AlterReplicaStateRequest,
+ callback: AlterReplicaStateResponseData => Unit): Unit
= {
+ val alterReplicaStateRequestDataData = alterReplicaStateRequest.data()
Review comment:
Should this be named `alterReplicaStateRequestData` instead?
##########
File path: core/src/main/scala/kafka/server/ReplicaManager.scala
##########
@@ -2281,4 +2288,31 @@ class ReplicaManager(val config: KafkaConfig,
}
}
}
+
+ def sentLogDirEvent(dir: String, topicPartitionsInDir:Seq[TopicPartition],
newState: Byte, reason: String): Unit = {
+ if (topicPartitionsInDir.nonEmpty) {
+ val logDirEventItem = AlterReplicaStateItem(topicPartitionsInDir.asJava,
newState, reason, handleAlterReplicaStateResponse)
+
+ logDirEventManager.handleAlterReplicaStateChanges(logDirEventItem)
+ } else {
+ info(s"Log dir: $dir contains none partitions, ignore log dir event")
+ }
+ }
+
+ /**
+ * Visible for testing
+ * @return true if we need not to retry, which means the response contains
no error or only UNKNOWN_TOPIC_OR_PARTITION error
Review comment:
This is incorrect, there's no return value
##########
File path: core/src/main/scala/kafka/server/KafkaServer.scala
##########
@@ -313,6 +315,21 @@ class KafkaServer(
}
alterIsrManager.start()
+ val alterReplicaStateChannelManager = new
BrokerToControllerChannelManagerImpl(
Review comment:
Same here, should we prefer `BrokerToControllerChannelManager`s `apply`?
##########
File path: clients/src/main/java/org/apache/kafka/common/protocol/Errors.java
##########
@@ -364,7 +365,8 @@
INCONSISTENT_TOPIC_ID(103, "The log's topic ID did not match the topic ID
in the request", InconsistentTopicIdException::new),
INCONSISTENT_CLUSTER_ID(104, "The clusterId in the request does not match
that found on the server", InconsistentClusterIdException::new),
TRANSACTIONAL_ID_NOT_FOUND(105, "The transactionalId could not be found",
TransactionalIdNotFoundException::new),
- FETCH_SESSION_TOPIC_ID_ERROR(106, "The fetch session encountered
inconsistent topic ID usage", FetchSessionTopicIdException::new);
+ FETCH_SESSION_TOPIC_ID_ERROR(106, "The fetch session encountered
inconsistent topic ID usage", FetchSessionTopicIdException::new),
+ UNKNOWN_REPLICA_STATE(107, "Replica state change only support
OfflineState, see ReplicaState.state ", UnknownReplicaStateException::new);
Review comment:
```suggestion
UNKNOWN_REPLICA_STATE(107, "Replica state change only supports
OfflineState, see ReplicaState.state ", UnknownReplicaStateException::new);
```
##########
File path: core/src/main/scala/kafka/server/BrokerServer.scala
##########
@@ -255,6 +257,21 @@ class BrokerServer(
)
alterIsrManager.start()
+ val alterReplicaStateChannelManager = new
BrokerToControllerChannelManagerImpl(
Review comment:
Everywhere else we seem to prefer `BrokerToControllerChannelManager`'s
`apply` instead of directly invoking `BrokerToControllerChannelManagerImpl`'s
constructor, should we stick to the same pattern?
##########
File path: core/src/main/scala/kafka/server/KafkaServer.scala
##########
@@ -313,6 +315,21 @@ class KafkaServer(
}
alterIsrManager.start()
+ val alterReplicaStateChannelManager = new
BrokerToControllerChannelManagerImpl(
+ controllerNodeProvider = MetadataCacheControllerNodeProvider(config,
metadataCache),
+ time = time,
+ metrics = metrics,
+ config = config,
+ channelName = "alterReplicaStateChannel",
+ threadNamePrefix = threadNamePrefix,
+ retryTimeoutMs = Long.MaxValue)
+ if (config.interBrokerProtocolVersion >= kafka.api.KAFKA_3_0_IV1) {
Review comment:
Why not `config.interBrokerProtocolVersion.isAlterReplicaStateSupported`?
##########
File path: core/src/main/scala/kafka/server/KafkaApis.scala
##########
@@ -3276,6 +3277,23 @@ class KafkaApis(val requestChannel: RequestChannel,
}
}
+ def handleAlterReplicaStateRequest(request: RequestChannel.Request): Unit = {
+ val zkSupport =
metadataSupport.requireZkOrThrow(KafkaApis.shouldNeverReceive(request))
Review comment:
Very likely a silly question, but this seems to be the only handler for
`AlterReplicaStateRequest` and I'm struggling to understand - how will this
work in KRaft, don't we need ControllerApis counterpart?
##########
File path: core/src/main/scala/kafka/server/LogDirEventManagerImpl.scala
##########
@@ -0,0 +1,198 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package kafka.server
+
+import java.util
+import java.util.concurrent.atomic.{AtomicBoolean, AtomicLong}
+import java.util.concurrent.{LinkedBlockingQueue, TimeUnit}
+
+import kafka.metrics.KafkaMetricsGroup
+import kafka.utils.{Logging, Scheduler}
+import org.apache.kafka.clients.ClientResponse
+import org.apache.kafka.common.TopicPartition
+import org.apache.kafka.common.message.{AlterReplicaStateRequestData,
AlterReplicaStateResponseData}
+import org.apache.kafka.common.protocol.Errors
+import org.apache.kafka.common.requests.{AlterReplicaStateRequest,
AlterReplicaStateResponse}
+import org.apache.kafka.common.utils.Time
+
+import scala.collection.mutable
+import scala.jdk.CollectionConverters._
+
+/**
+ * Handles the sending of AlterReplicaState requests to the controller when
LogDirFailure.
+ */
+abstract class LogDirEventManager {
+
+ def start(): Unit
+
+ def handleAlterReplicaStateChanges(logDirEventItem: AlterReplicaStateItem):
Unit
+
+ def pendingAlterReplicaStateItemCount(): Int
+
+ def shutdown(): Unit
+}
+
+case class AlterReplicaStateItem(topicPartitions: util.List[TopicPartition],
+ newState: Byte,
+ reason: String,
+ callback: Either[Errors, TopicPartition] =>
Unit)
+
+class LogDirEventManagerImpl(val controllerChannelManager:
BrokerToControllerChannelManager,
+ val scheduler: Scheduler,
+ val time: Time,
+ val brokerId: Int,
+ val brokerEpochSupplier: () => Long) extends
LogDirEventManager with Logging with KafkaMetricsGroup {
+
+ // Used to allow only one in-flight request at a time
+ private val inflightRequest: AtomicBoolean = new AtomicBoolean(false)
+
+ private val pendingReplicaStateUpdates:
LinkedBlockingQueue[AlterReplicaStateItem] = new
LinkedBlockingQueue[AlterReplicaStateItem]()
+
+ private val lastSentMs = new AtomicLong(0)
+
+ def start(): Unit = {
+ scheduler.schedule("send-alter-replica-state",
propagateReplicaStateChanges, 50, 50, TimeUnit.MILLISECONDS)
+ }
+
+ override def pendingAlterReplicaStateItemCount(): Int =
pendingReplicaStateUpdates.size()
+
+ private def propagateReplicaStateChanges(): Unit = {
+ if (!pendingReplicaStateUpdates.isEmpty &&
inflightRequest.compareAndSet(false, true)) {
+ // Copy current unsent items and remove from the queue, will be inserted
if failed
+ val inflightAlterIsrItem = pendingReplicaStateUpdates.poll()
+
+ lastSentMs.set(time.milliseconds())
+ sendRequest(inflightAlterIsrItem)
+ }
+ }
+
+ def handleAlterReplicaStateChanges(logDirEventItem: AlterReplicaStateItem):
Unit = {
+ pendingReplicaStateUpdates.put(logDirEventItem)
+ }
+
+ def sendRequest(logDirEventItem: AlterReplicaStateItem): Unit = {
+
+ val message = buildRequest(logDirEventItem)
+
+ debug(s"Sending AlterReplicaState to controller $message")
+ controllerChannelManager.sendRequest(new
AlterReplicaStateRequest.Builder(message),
+ new ControllerRequestCompletionHandler {
+ override def onComplete(response: ClientResponse): Unit = {
+ try {
+ val body =
response.responseBody().asInstanceOf[AlterReplicaStateResponse]
+ handleAlterReplicaStateResponse(body, message.brokerEpoch,
logDirEventItem)
+ } finally {
+ inflightRequest.set(false)
+ }
+ }
+
+ override def onTimeout(): Unit = {
+ throw new IllegalStateException("Encountered unexpected timeout when
sending AlterIsr to the controller")
+ }
+ })
+ }
+
+ private def buildRequest(logDirEventItem: AlterReplicaStateItem):
AlterReplicaStateRequestData = {
+ val message = new AlterReplicaStateRequestData()
+ .setBrokerId(brokerId)
+ .setBrokerEpoch(brokerEpochSupplier.apply())
+ .setNewState(logDirEventItem.newState)
+ .setReason(logDirEventItem.reason)
+ .setTopics(new java.util.ArrayList())
+
+ logDirEventItem.topicPartitions.asScala.groupBy(_.topic).foreach(entry => {
+ val topicPart = new AlterReplicaStateRequestData.TopicData()
+ .setName(entry._1)
+ .setPartitions(new java.util.ArrayList())
+ message.topics().add(topicPart)
+ entry._2.foreach(item => {
+ topicPart.partitions().add(new
AlterReplicaStateRequestData.PartitionData()
+ .setPartitionIndex(item.partition)
+ )
+ })
+ })
+ message
+ }
+
+ private def handleAlterReplicaStateResponse(alterReplicaStateResponse:
AlterReplicaStateResponse,
+ sentBrokerEpoch: Long,
+ logDirEventItem:
AlterReplicaStateItem): Unit = {
+ val data: AlterReplicaStateResponseData = alterReplicaStateResponse.data
+
+ Errors.forCode(data.errorCode) match {
+ case Errors.STALE_BROKER_EPOCH =>
+ warn(s"Broker had a stale broker epoch ($sentBrokerEpoch), broker
could have been repaired and restarted, ignore")
+ pendingReplicaStateUpdates.put(logDirEventItem)
+
+ case Errors.NOT_CONTROLLER =>
+ warn(s"Remote broker is not controller, ignore")
+ pendingReplicaStateUpdates.put(logDirEventItem)
+
+ case Errors.CLUSTER_AUTHORIZATION_FAILED =>
+ val exception = Errors.CLUSTER_AUTHORIZATION_FAILED.exception("Broker
is not authorized to send AlterReplicaState to controller")
+ error(s"Broker is not authorized to send AlterReplicaState to
controller", exception)
+ pendingReplicaStateUpdates.put(logDirEventItem)
+
+ case Errors.UNKNOWN_REPLICA_STATE =>
+ val exception =
Errors.CLUSTER_AUTHORIZATION_FAILED.exception("ReplicaStateChange failed with
an unknown replica state")
+ error(s"Broker is not authorized to send AlterReplicaState to
controller", exception)
+ pendingReplicaStateUpdates.put(logDirEventItem)
+
+ case Errors.NONE =>
+ // success is a flag to indicate whether all the partitions had
successfully alter state
+ val failedPartitions = new util.ArrayList[TopicPartition]()
+ // Collect partition-level responses to pass to the callbacks
+ val partitionResponses: mutable.Map[TopicPartition, Either[Errors,
TopicPartition]] =
+ new mutable.HashMap[TopicPartition, Either[Errors, TopicPartition]]()
+ data.topics.forEach { topic =>
+ topic.partitions().forEach(partition => {
+ val tp = new TopicPartition(topic.name, partition.partitionIndex)
+ val error = Errors.forCode(partition.errorCode())
+ debug(s"Controller successfully handled AlterReplicaState request
for $tp: $partition")
+ if (error == Errors.NONE) {
+ partitionResponses(tp) = Right(tp)
+ } else {
+ failedPartitions.add(new TopicPartition(topic.name(),
partition.partitionIndex()))
+ partitionResponses(tp) = Left(error)
+ }
+ })
+ }
+
+ // Iterate across the items we sent rather than what we received to
ensure we run the callback even if a
+ // partition was somehow erroneously excluded from the response.
Review comment:
Is it worth also checking that there aren't partitions in the response
that weren't part of the request?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
