New to Akka? Start with the Akka SDK.

Migration Guide 2.10.x to 2.11.x

Jackson 2.21

Akka Serialization Jackson now depends on the Jackson 2.21 long-term-support line. Akka 2.10 used Jackson 2.18 which will not get further updates past January 2027.

Removed deprecated APIs

APIs deprecated in Akka 2.10.0 or earlier have been removed. The sections below list the removed APIs and their replacements.

akka-actor-testkit-typed

LoggingTestKit.intercept (scaladsl): use expect instead.

akka-actor-typed

akka.actor.typed.scaladsl.LoggerOps: not needed in Scala 2.13 and later, use the org.slf4j.Logger methods directly.
TimerScheduler.startPeriodicTimer (scaladsl and javadsl): use startTimerWithFixedDelay or startTimerAtFixedRate instead. In scaladsl, startPeriodicTimer had the same semantics as startTimerAtFixedRate. In javadsl, it had the same semantics as startTimerWithFixedDelay, which is often preferred.

akka-cluster

UniqueAddress constructor and UniqueAddress.apply with an Int uid, UniqueAddress.uid and UniqueAddress.copy: use the Long uid constructor or apply, and longUid.
ClusterEvent.CurrentClusterState constructors without unreachableDataCenters and memberTombstones: use the main constructor, or CurrentClusterState.apply, which keeps the default arguments.

akka-cluster-sharding

The deprecated ClusterShardingSettings and ClusterShardingSettings.TuningParameters constructors: use ClusterShardingSettings(system) or ClusterShardingSettings.create(system) with the with* methods, or the full constructor.
ClusterShardingSettings.passivateIdleEntityAfter and withPassivateIdleAfter: use withPassivationStrategy and passivationStrategySettings, see Automatic Passivation.
ClusterShardingHealthCheckSettings(names, timeout): use the full constructor.

akka-cluster-sharding-typed

The deprecated ClusterShardingSettings constructors: use ClusterShardingSettings(system) or ClusterShardingSettings.create(system) with the with* methods, or the full constructor.
ClusterShardingSettings.passivateIdleEntityAfter and withPassivateIdleEntityAfter: use withPassivationStrategy and passivationStrategySettings, see Automatic Passivation.
ShardingProducerController.Settings.resendFirsUnconfirmedIdleTimeout and withResendFirsUnconfirmedIdleTimeout: use the correctly spelled resendFirstUnconfirmedIdleTimeout and withResendFirstUnconfirmedIdleTimeout.

akka-cluster-tools

DistributedPubSubSettings constructor without sendToDeadLettersWhenNoSubscribers: use the constructor with sendToDeadLettersWhenNoSubscribers, or DistributedPubSubSettings(config).

akka-cluster-typed

ClusterSingletonSettings and ClusterSingletonManagerSettings constructors without leaseSettings: use the constructor with leaseSettings.

akka-distributed-data

ReplicatorSettings constructors with fewer parameters than the full constructor: use the full constructor, ReplicatorSettings(config) or the with* methods.
ORMap.updated(Cluster, key, initial, modify) (Java API): use ORMap.update(SelfUniqueAddress, key, initial, modify).
GSet.copy$default$1, a binary compatibility stub in the Scala 2 sources only.

akka-multi-node-testkit

TestConductor.throttle: throttle was not implemented, use blackhole and passThrough instead.

akka-persistence

The LevelDB journal (akka.persistence.journal.leveldb, LeveldbJournal) and the shared LevelDB journal (akka.persistence.journal.leveldb-shared, SharedLeveldbJournal, SharedLeveldbStore): use another journal implementation, such as Akka Persistence R2DBC. For tests, use the akka.persistence.journal.inmem journal, in combination with the journal proxy when several actor systems share one journal. The akka.persistence.journal.leveldb and akka.persistence.journal.leveldb-shared configuration sections no longer exist.
DurableStateUpdateStore.deleteObject(persistenceId) in akka.persistence.state.scaladsl and akka.persistence.state.javadsl: use the deleteObject overload with revision instead. Durable state store plugins must remove their implementation of the overload without revision.

akka-persistence-query

The LevelDB read journal (akka.persistence.query.journal.leveldb, LeveldbReadJournal in scaladsl and javadsl, LeveldbReadJournalProvider): use another journal and query implementation, such as Akka Persistence R2DBC. The akka.persistence.query.journal.leveldb configuration section no longer exists.
EventEnvelope(offset, persistenceId, sequenceNr, event) and new EventEnvelope(offset, persistenceId, sequenceNr, event) in akka.persistence.query: use the overloads with a timestamp parameter. The EventEnvelope companion object no longer extends Function4.

akka-persistence-typed

snapshotSelectionCriteria in the javadsl EventSourcedBehavior, EventSourcedOnCommandBehavior and EventSourcedOnCommandWithReplyBehavior: override recovery instead and return Recovery.withSnapshotSelectionCriteria(...). Overriding snapshotSelectionCriteria had no effect since Akka 2.6.5.
EventSourcedBehavior.withSnapshotSelectionCriteria in scaladsl: use withRecovery(Recovery.withSnapshotSelectionCriteria(...)).

akka-remote

AddressUidExtension: use ExtendedActorSystem.uid instead. In general akka.cluster.UniqueAddress is a better source for getting the uid. With typed Cluster that is accessible via selfMember and with classic Cluster via selfUniqueAddress.
akka.remote.serialization.LongSerializer, IntSerializer, StringSerializer and ByteStringSerializer: moved to akka.serialization in akka-actor. The serializer ids 18 to 21 are unchanged and used by the akka-actor serializers, so persisted data still deserializes.

akka-stream

ActorMaterializer: Materializer has all the methods ActorMaterializer had. Use the system materializer: Materializer.matFromSystem provides it implicitly when an implicit ActorSystem, or another ClassicActorSystemProvider, is in scope. Replace ActorMaterializer(...)(context) in an actor with Materializer(context) or Materializer.createMaterializer(context), and ActorMaterializer(...)(system) with Materializer(system), to create a separate materializer.
The ActorMaterializerSettings, IOSettings and StreamRefSettings factories and with* methods, and Materializer.settings: configure streams with the akka.stream.materializer config, or per stream with ActorAttributes, TcpAttributes and StreamRefAttributes. Operators read the settings from their attributes.
The settings accessors are no longer public. Read the attribute that replaces each: ActorMaterializerSettings.dispatcher: ActorAttributes.dispatcher; initialInputBufferSize and maxInputBufferSize: Attributes.inputBuffer; supervisionDecider: ActorAttributes.supervisionStrategy; debugLogging: ActorAttributes.debugLogging; outputBurstLimit: ActorAttributes.outputBurstLimit; fuzzingMode: ActorAttributes.fuzzingMode; maxFixedBufferSize: ActorAttributes.maxFixedBufferSize; syncProcessingLimit: ActorAttributes.syncProcessingLimit; blockingIoDispatcher: the ActorAttributes.IODispatcher attribute, which selects the akka.stream.materializer.blocking-io-dispatcher config; autoFusing: no replacement, it had no effect. IOSettings.tcpWriteBufferSize: TcpAttributes.tcpWriteBufferSize. StreamSubscriptionTimeoutSettings.mode and timeout: ActorAttributes.streamSubscriptionTimeout. StreamRefSettings.bufferCapacity, demandRedeliveryInterval, subscriptionTimeout and finalTerminationSignalDeadline: the StreamRefAttributes methods with the same names.
IOResult.status, wasSuccessful, getError and createFailed, and AbruptIOTerminationException: IO operators fail the materialized value with IOOperationIncompleteException. FileIO.fromPath now also fails the materialized value when the file can’t be read, instead of completing it with a failed status.
The copy, unapply and Product methods of Attributes: use the explicit methods on Attributes. Attributes no longer extends Product.
Source.fromFuture, fromCompletionStage, fromFutureSource and fromSourceCompletionStage: use Source.future, Source.completionStage, Source.futureSource and Source.completionStageSource.
Source.lazily and Source.lazilyAsync: use Source.lazySource, Source.lazyFuture or Source.lazyCompletionStage.
Sink.lazyInit and Sink.lazyInitAsync: use Sink.lazyFutureSink or Sink.lazyCompletionStageSink. Flow.lazyInit and Flow.lazyInitAsync: use Flow.lazyFutureFlow or Flow.lazyCompletionStageFlow.
Source.setup, Flow.setup and Sink.setup: use fromMaterializer.
Source.actorRef(bufferSize, overflowStrategy) and Sink.actorRef(ref, onCompleteMessage): use the overloads that take completion and failure matchers, or an on-complete and an on-failure message. Source.actorRefWithAck and Sink.actorRefWithAck: use actorRefWithBackpressure.
FileIO.fromFile and FileIO.toFile: use FileIO.fromPath and FileIO.toPath.
The akka.stream.javadsl.Tcp.bind and outgoingConnection overloads that take a scala.concurrent.duration.Duration: use the overloads that take a java.time.Duration.
The RestartSource, RestartFlow and RestartSink overloads without RestartSettings: use the overloads that take RestartSettings.
TimerGraphStageLogic.schedulePeriodically, schedulePeriodicallyWithInitialDelay and Materializer.schedulePeriodically: use scheduleWithFixedDelay or scheduleAtFixedRate.
The BindFailedException object: match on the BindFailedException class instead.

akka-stream-testkit

TestSink.probe and TestSource.probe (scaladsl): use TestSink() and TestSource() with an implicit ClassicActorSystemProvider instead.
TestSink.probe(system) and TestSource.probe(system) (javadsl): use TestSink.create(system) and TestSource.create(system) with a ClassicActorSystemProvider instead.

akka-stream-typed

ActorSource.actorRefWithAck (scaladsl and javadsl): use ActorSource.actorRefWithBackpressure instead.

akka-testkit

akka.testkit.javadsl.TestKit overloads of expectMsgPF, expectTerminated, fishForMessage and fishForSpecificMessage taking a scala.concurrent.duration.Duration: use the overloads that accept java.time.Duration instead.