PANTHEON.tech has enabled OpenDaylight to migrate to the current version of AKKA, 2.6.x. Today, we will review recent changes to AKKA, which is the heart of OpenDaylight’s Clustering functionality.
As the largest committer to the OpenDaylight source-code, PANTHEON.tech will regularly keep you updated and posted about our efforts in projects surrounding OpenDaylight.
Released in November 2019, added many improvements, including better documentation, and introduced a lot of new features. In case you are new to AKKA, here is a description from the official website:
[It is] a set of open-source libraries for designing scalable, resilient systems that span processor cores and networks. [AKKA] allows you to focus on meeting business needs instead of writing low-level code to provide reliable behavior, fault tolerance, and high performance.
OpenDaylight Migrates to AKKA 2.6
The most important features of AKKA 2.6 are shortly described below; for the full list, please see the release notes.
AKKA Typed is the new typed Actor API. It was declared ready for production use since AKKA 2.5.22, and since 2.6.0 it’s the officially recommended Actor API for new projects. Also, the documentation was significantly improved.
In a typed API, each actor declares an acceptable message type, and only messages of this type can be sent to the actor. This is enforced by the system.
For untyped extensions, seamless access is allowed.
The classic APIs for modules, such as Persistence, Cluster Sharding and Distributed Data, are still fully supported, so the existing applications can continue to use those.
Artery is a reimplementation of the old remoting module, aimed at improving performance and stability, based on Aeron (UDP) and AKKA Streams TCP/TLS, instead of Netty TCP.
Artery provides more stability and better performance for systems using AKKA Cluster. Classic remoting is deprecated but still can be used. More information is provided in the migration guide.
AKKA 2.6 includes a new Jackson-based serializer, supporting both JSON and CBOR formats. This is the recommended serializer for applications using AKKA 2.6. Java serialization is disabled by default.
With AKKA Typed, actors on any node in the cluster can subscribe to specific topics. The message published to one of these topics will be delivered to all subscribed actors. This feature also works in a non-cluster setting.
Passivation in Cluster
With the Cluster passivation feature, you may stop persistent entities that are not used to reduce memory consumption by defining a timeout for message receiving. Passivation timeout can be pre-configured in cluster settings or set explicitly for an entity.
Cluster: External shard allocation
A possibility to use a new alternative external shard allocation strategy is provided, which allows explicit control over the allocation of shards. This covers use cases such as Kafka Partition consumption matching with shard locations to avoid network hops.
Sharded Daemon Process
Sharded Daemon Process allows specific actors in a cluster, to keep it alive and balanced. For rebalancing, an actor can be stopped and then started on a new node. This feature can be useful in case the data processing workload needs to be split across a set number of workers.
Originally published at https://pantheon.tech on February 1, 2021.