Memberlist

memberlist is a rust crate that manages cluster membership and member failure detection using a gossip based protocol.

The use cases for such a library are far-reaching: all distributed systems require membership, and memberlist is a re-usable solution to managing cluster membership and node failure detection.

memberlist is eventually consistent but converges quickly on average. The speed at which it converges can be heavily tuned via various knobs on the protocol. Node failures are detected and network partitions are partially tolerated by attempting to communicate to potentially dead nodes through multiple routes.

memberlist is WASM/WASI friendly, all crates can be compiled to wasm-wasi and wasm-unknown-unknown (need to configure the crate features).

Unlike the original Go implementation, Rust's memberlist use highly generic and layered architecture, users can easily implement a component by themselves and plug it to the memberlist. Users can even custom their own Id and Address.

Here are the layers:

• Transport Layer

  By default, Rust's memberlist provides two kinds of transport -- QuicTransport and NetTransport.

  • Runtime Layer

    Async runtime agnostic are provided by agnostic's Runtime trait, tokio, async-std and smol are supported by default. Users can implement their own Runtime and plug it into the memberlist.

  • Address Resolver Layer

    The address resolver layer is supported by nodecraft's AddressResolver trait.

  • Serialize/Deserilize Layer

    By default, Rust's memberlist is using length-prefix encoding (Lpe) to serialize/deserialize messages to bytes or visa-vise. The implemention of Lpe tries the best to avoid reallocating when doing the serialize/deserialize.

    But, users can use any other serialize/deserialize framework by implementing Wire trait.

  • NetTransport

    Three kinds of different builtin stream layers for NetTransport:

    • Tcp: based on TCP and UDP
    • Tls: based on rustls and UDP
    • NativeTls: based on native-tls and UDP

  • QuicTransport

    QUIC transport is an experimental transport implementation, it is well tested but still experimental.

    Two kinds of different builtin stream layers for QuicTransport:

    • Quinn: based on quinn
    • S2n: based on s2n-quic

  Users can still implement their own stream layer for different kinds of transport implementations.

• Delegate Layer

  This layer is used as a reactor for different kinds of messages.

  • Delegate

    Delegate is the trait that clients must implement if they want to hook into the gossip layer of Memberlist. All the methods must be thread-safe, as they can and generally will be called concurrently.

    Here are the sub delegate traits:

    • AliveDelegate

      Used to involve a client in processing a node "alive" message. When a node joins, either through a packet gossip or promised push/pull, we update the state of that node via an alive message. This can be used to filter a node out and prevent it from being considered a peer using application specific logic.

    • ConflictDelegate

      Used to inform a client that a node has attempted to join which would result in a name conflict. This happens if two clients are configured with the same name but different addresses.

    • EventDelegate

      A simpler delegate that is used only to receive notifications about members joining and leaving. The methods in this delegate may be called by multiple threads, but never concurrently. This allows you to reason about ordering.

    • MergeDelegate

      Used to involve a client in a potential cluster merge operation. Namely, when a node does a promised push/pull (as part of a join), the delegate is involved and allowed to cancel the join based on custom logic. The merge delegate is NOT invoked as part of the push-pull anti-entropy.

    • NodeDelegate

      Used to manage node related events. e.g. metadata

    • PingDelegate

      Used to notify an observer how long it took for a ping message to complete a round trip. It can also be used for writing arbitrary byte slices into ack messages. Note that in order to be meaningful for RTT estimates, this delegate does not apply to indirect pings, nor fallback pings sent over promised connection.

  • CompositeDelegate

    CompositeDelegate is a helpful struct to split the Delegate into multiple small delegates, so that users do not need to implement full Delegate when they only want to custom some methods in the Delegate.

memberlist is based on "SWIM: Scalable Weakly-consistent Infection-style Process Group Membership Protocol". However, Hashicorp developers extends the protocol in a number of ways:

Several extensions are made to increase propagation speed and convergence rate. Another set of extensions, that Hashicorp developers call Lifeguard, are made to make memberlist more robust in the presence of slow message processing (due to factors such as CPU starvation, and network delay or loss). For details on all of these extensions, please read Hashicorp's paper "Lifeguard : SWIM-ing with Situational Awareness", along with the memberlist source.

[dependencies]
memberlist = "0.1"

• Does Rust's memberlist implemenetation compatible to Go's memberlist?

  No but yes! By default, it is not compatible. But the secret is the serialize/deserilize layer, Go's memberlist use the msgpack as the serialization/deserialization framework, so in theory, if you can implement a Wire trait which compat to Go's memberlist, then it becomes compatible.

• If Go's memberlist adds more functionalities, will this project also support?

  Yes! And this project may also add more functionalities whereas the Go's memberlist does not have. e.g. wasmer support, bindings to other languages and etc.

• agnostic: helps you to develop runtime agnostic crates
• nodecraft: crafting seamless node operations for distributed systems, which provides foundational traits for node identification and address resolution.
• transformable: transform its representation between structured and byte form.
• peekable: peekable reader and async reader

memberlist is under the terms of the MPL-2.0 license.

See LICENSE for details.

Copyright (c) 2024 Al Liu.

Copyright (c) 2013 HashiCorp, Inc.