KAFKA-14840: Support for snapshots during ZK migration (apache#13461)

This patch adds support for handling metadata snapshots while in dual-write mode. Prior to this change, if the active
controller loaded a snapshot, it would get out of sync with the ZK state.

In order to reconcile the snapshot state with ZK, several methods were added to scan through the metadata in ZK to
compute differences with the MetadataImage. Since this introduced a lot of code, I opted to split out a lot of methods
from ZkMigrationClient into their own client interfaces, such as TopicMigrationClient, ConfigMigrationClient, and
AclMigrationClient. Each of these has some iterator method that lets the caller examine the ZK state in a single pass
and without using too much memory.

Reviewers: Colin P. McCabe <cmccabe@apache.org>, Luke Chen <showuon@gmail.com>

checkstyle/suppressions.xml

@@ -300,7 +300,7 @@
 
     <!-- metadata -->
     <suppress checks="ClassDataAbstractionCoupling"
-              files="(QuorumController|QuorumControllerTest|ReplicationControlManager|ReplicationControlManagerTest|ClusterControlManagerTest).java"/>
+              files="(QuorumController|QuorumControllerTest|ReplicationControlManager|ReplicationControlManagerTest|ClusterControlManagerTest|KRaftMigrationDriverTest).java"/>
     <suppress checks="ClassFanOutComplexity"
               files="(QuorumController|QuorumControllerTest|ReplicationControlManager|ReplicationControlManagerTest).java"/>
     <suppress checks="(ParameterNumber|ClassDataAbstractionCoupling)"

core/src/main/scala/kafka/server/ControllerServer.scala

@@ -259,7 +259,7 @@ class ControllerServer(
             config.passwordEncoderIterations)
           case None => PasswordEncoder.noop()
         }
-        val migrationClient = new ZkMigrationClient(zkClient, zkConfigEncoder)
+        val migrationClient = ZkMigrationClient(zkClient, zkConfigEncoder)
         val propagator: LegacyPropagator = new MigrationPropagator(config.nodeId, config)
         val migrationDriver = new KRaftMigrationDriver(
           config.nodeId,

core/src/main/scala/kafka/zk/migration/ZkAclMigrationClient.scala

@@ -0,0 +1,117 @@
+/*
+ * 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.zk.migration
+
+import kafka.security.authorizer.AclAuthorizer.{ResourceOrdering, VersionedAcls}
+import kafka.security.authorizer.{AclAuthorizer, AclEntry}
+import kafka.utils.Logging
+import kafka.zk.ZkMigrationClient.{logAndRethrow, wrapZkException}
+import kafka.zk.{KafkaZkClient, ResourceZNode, ZkAclStore, ZkVersion}
+import kafka.zookeeper.{CreateRequest, DeleteRequest, SetDataRequest}
+import org.apache.kafka.common.acl.AccessControlEntry
+import org.apache.kafka.common.resource.ResourcePattern
+import org.apache.kafka.metadata.migration.{AclMigrationClient, MigrationClientException, ZkMigrationLeadershipState}
+import org.apache.zookeeper.CreateMode
+import org.apache.zookeeper.KeeperException.Code
+
+import java.util
+import java.util.function.BiConsumer
+import scala.jdk.CollectionConverters._
+
+class ZkAclMigrationClient(
+  zkClient: KafkaZkClient
+) extends AclMigrationClient with Logging {
+
+  private def aclChangeNotificationRequest(resourcePattern: ResourcePattern): CreateRequest = {
+    // ZK broker needs the ACL change notification znode to be updated in order to process the new ACLs
+    val aclChange = ZkAclStore(resourcePattern.patternType).changeStore.createChangeNode(resourcePattern)
+    CreateRequest(aclChange.path, aclChange.bytes, zkClient.defaultAcls(aclChange.path), CreateMode.PERSISTENT_SEQUENTIAL)
+  }
+
+  private def tryWriteAcls(
+    resourcePattern: ResourcePattern,
+    aclEntries: Set[AclEntry],
+    create: Boolean,
+    state: ZkMigrationLeadershipState
+  ): Option[ZkMigrationLeadershipState] = wrapZkException {
+    val aclData = ResourceZNode.encode(aclEntries)
+
+    val request = if (create) {
+      val path = ResourceZNode.path(resourcePattern)
+      CreateRequest(path, aclData, zkClient.defaultAcls(path), CreateMode.PERSISTENT)
+    } else {
+      SetDataRequest(ResourceZNode.path(resourcePattern), aclData, ZkVersion.MatchAnyVersion)
+    }
+
+    val (migrationZkVersion, responses) = zkClient.retryMigrationRequestsUntilConnected(Seq(request), state)
+    if (responses.head.resultCode.equals(Code.NONODE)) {
+      // Need to call this method again with create=true
+      None
+    } else {
+      // Write the ACL notification outside of a metadata multi-op
+      zkClient.retryRequestUntilConnected(aclChangeNotificationRequest(resourcePattern))
+      Some(state.withMigrationZkVersion(migrationZkVersion))
+    }
+  }
+
+  override def writeResourceAcls(
+    resourcePattern: ResourcePattern,
+    aclsToWrite: util.Collection[AccessControlEntry],
+    state: ZkMigrationLeadershipState
+  ): ZkMigrationLeadershipState = {
+    val acls = aclsToWrite.asScala.map(new AclEntry(_)).toSet
+    tryWriteAcls(resourcePattern, acls, create = false, state) match {
+      case Some(newState) => newState
+      case None => tryWriteAcls(resourcePattern, acls, create = true, state) match {
+        case Some(newState) => newState
+        case None => throw new MigrationClientException(s"Could not write ACLs for resource pattern $resourcePattern")
+      }
+    }
+  }
+
+  override def deleteResource(
+    resourcePattern: ResourcePattern,
+    state: ZkMigrationLeadershipState
+  ): ZkMigrationLeadershipState = {
+    val request = DeleteRequest(ResourceZNode.path(resourcePattern), ZkVersion.MatchAnyVersion)
+    val (migrationZkVersion, responses) = zkClient.retryMigrationRequestsUntilConnected(Seq(request), state)
+    if (responses.head.resultCode.equals(Code.OK) || responses.head.resultCode.equals(Code.NONODE)) {
+      // Write the ACL notification outside of a metadata multi-op
+      zkClient.retryRequestUntilConnected(aclChangeNotificationRequest(resourcePattern))
+      state.withMigrationZkVersion(migrationZkVersion)
+    } else {
+      throw new MigrationClientException(s"Could not delete ACL for resource pattern $resourcePattern")
+    }
+  }
+
+  override def iterateAcls(
+    aclConsumer: BiConsumer[ResourcePattern, util.Set[AccessControlEntry]]
+  ): Unit = {
+    // This is probably fairly inefficient, but it preserves the semantics from AclAuthorizer (which is non-trivial)
+    var allAcls = new scala.collection.immutable.TreeMap[ResourcePattern, VersionedAcls]()(new ResourceOrdering)
+    def updateAcls(resourcePattern: ResourcePattern, versionedAcls: VersionedAcls): Unit = {
+      allAcls = allAcls.updated(resourcePattern, versionedAcls)
+    }
+    AclAuthorizer.loadAllAcls(zkClient, this, updateAcls)
+    allAcls.foreach { case (resourcePattern, versionedAcls) =>
+      logAndRethrow(this, s"Error in ACL consumer. Resource was $resourcePattern.") {
+        aclConsumer.accept(resourcePattern, versionedAcls.acls.map(_.ace).asJava)
+      }
+    }
+  }
+}