GossNet.Discovery.Redis 0.10.0

GossNet.Discovery.Redis

Redis-based node discovery for GossNet.Protocol.

Unlike the read-only providers, this one registers the node. Consul has an agent and EC2 has instance tags, but Redis knows nothing until something writes to it — so each node heartbeats itself into a shared sorted set and reads the others out of it.

Installation

dotnet add package GossNet.Discovery.Redis

Usage

using GossNet.Discovery.Redis;
using GossNet.Protocol;

var redisOptions = new RedisDiscoveryOptions
{
    ConnectionString = "localhost:6379",
    Key = "gossnet:members"
};

var configuration = new GossNetConfiguration
{
    Hostname = "10.0.0.4",
    Port = 9055,
    DiscoveryProviderFactory = cfg => new RedisNodeDiscovery(cfg, redisOptions)
};

await using var node = new GossNetNode<MyMessage>(configuration);
node.Start();

Hostname and Port are what this node registers, so they must be reachable by the others.

If your application already talks to Redis, share the multiplexer rather than opening a second connection — it is expensive and designed to be shared:

DiscoveryProviderFactory = cfg => new RedisNodeDiscovery(
    cfg, redisOptions, registry: new RedisRegistry(existingMultiplexer))

A registry you supply is left open when discovery is disposed; one built from ConnectionString is closed.

How membership is stored

One sorted set. The member is host:port, the score is the epoch milliseconds of its last heartbeat:

ZADD gossnet:members <now-ms> "10.0.0.4:9055"      # heartbeat, every HeartbeatInterval
ZRANGEBYSCORE gossnet:members <cutoff> +inf        # who is alive
ZREM gossnet:members "10.0.0.4:9055"               # clean shutdown

A single key and a single range query, with no per-key expiry and no SCAN — both of which scale badly on a shared instance. Entries older than twice the timeout are pruned during heartbeats, so a node that never comes back does not accumulate forever.

Disposing the provider removes this node's entry, so a clean shutdown is noticed immediately rather than after RegistrationTimeout.

Options

¹ Not required when you pass a RedisRegistry built from an existing multiplexer.

Keep RegistrationTimeout at several heartbeat intervals. A node that misses one because of a GC pause or a slow round trip has not left the cluster.

Use a different Key per cluster sharing a Redis instance, the way you would a different Consul service name.

A failed read throws NodeDiscoveryException rather than returning an empty list, so an unreachable Redis is never mistaken for a cluster of one. A failed heartbeat is only logged — the timeout is several intervals wide, so one miss is survivable.

License

MIT