Cluster inspection

The Server facet offers typed, read-only Cluster inspection. Use it to see a node's view of the topology, slots, links, and slot statistics. An ordinary call reports one node's view. It does not update the client's slot map, merge a global topology, change membership, or enable Cluster mode. AllowAdmin is not required; the server still enforces ACLs.

API	Redis command	Minimum Redis version
ClusterInfoAsync	CLUSTER INFO	3.0
ClusterNodesAsync	CLUSTER NODES	3.0
ClusterShardsAsync	CLUSTER SHARDS	7.0
ClusterLinksAsync	CLUSTER LINKS	7.0
ClusterMyIdAsync	CLUSTER MYID	3.0
ClusterMyShardIdAsync	CLUSTER MYSHARDID	7.2
ClusterKeySlotAsync	CLUSTER KEYSLOT	3.0
ClusterCountKeysInSlotAsync	CLUSTER COUNTKEYSINSLOT	3.0
ClusterSlotStatsAsync	CLUSTER SLOT-STATS SLOTSRANGE	8.2
ClusterSlotStatsByMetricAsync	CLUSTER SLOT-STATS ORDERBY	8.2

Unsupported commands and standalone servers throw RespireServerException; no fake topology or empty success is returned. RESP2 and RESP3 are both supported.

RespireClusterInfo info = await redis.Server.ClusterInfoAsync();
ulong? currentEpoch = info.CurrentEpoch;
ulong? localEpoch = info.MyEpoch;
Console.WriteLine($"Current epoch: {currentEpoch}; local epoch: {localEpoch}");
RespireClusterShard[] shards = await redis.Server.ClusterShardsAsync();
foreach (var shard in shards)
{
    foreach (var member in shard.Nodes)
        Console.WriteLine($"{member.Id}: {member.Role}, {member.Health}, {member.Endpoint}");
}

Owned node views

RespireClusterInfo:

• Exposes common counters and keeps every field's original text in Attributes.
• CurrentEpoch and MyEpoch are ulong?, covering the full unsigned 64-bit range (including values above long.MaxValue); missing fields are null. Store them in ulong?, not long?. See the epoch specification.
• Other counters are nonnegative long? values.

RespireClusterNode (NODES) keeps the advertised address text, flags, primary ID, timing and epoch counters, link state, inclusive slot ranges, and importing/migrating annotations. Unknown trailing tokens go to AdditionalTokens. Unusable addresses such as :0@0 are kept for inspection.

Shard models (SHARDS) hold slot ranges and member nodes:

• Endpoint, IP, hostname, plain port, and TLS port are separate. Null, empty, and ? endpoints are kept; port zero means unavailable.
• The client does not turn advertised values into reachable endpoints or guess missing addresses.
• Role and health strings keep future values.

Link models (CLUSTER LINKS) describe Cluster bus connections, not the client's: direction, peer ID, creation time (Unix milliseconds), event flags, and buffer sizes.

All models:

• Unknown fields go to AdditionalFields as RespireResult values. Disposing them is optional and invalidates their element views. Other models need no disposal.
• Results stay valid after the reply and client are disposed.
• Each call returns its own snapshot. Arrays are caller-owned and mutable. Record equality compares array references, so compare elements yourself when comparing snapshots.

Keys, slots, and statistics

• ClusterKeySlotAsync applies the view's key prefix and keeps binary key bytes. The node calculates the slot; the call is not routed to the key's owner.
• Numeric slot arguments must be 0–16383. Ranges are inclusive and cannot be reversed.
• ClusterCountKeysInSlotAsync counts keys on the queried node, even if another node owns the slot. Use the all-node variant to compare nodes. Summing primary and replica counts double-counts data.

int slot = await redis.Server.ClusterKeySlotAsync("orders:42");
RespireClusterSlotStats[] local = await redis.Server.ClusterSlotStatsAsync(slot, slot);
RespireClusterSlotStats[] busiest = await redis.Server.ClusterSlotStatsByMetricAsync(
    RespireClusterSlotMetric.KeyCount, limit: 10, descending: true);

CLUSTER SLOT-STATS:

• Returns slots in the queried node's shard. Range results keep server order.
• Metric results default to descending; limit is 1 to 16384.
• KeyCount is always present. Other metrics are null when disabled in server configuration. Ordering by a disabled metric returns the server error.
• These APIs do not enable statistics collection.

Results from every node

Every method has an OnAllNodesAsync counterpart with the same arguments. It returns a RespireServerResult<T> per endpoint with an owned value or the original failure.

• Discovery includes replicas and slotless members, and needs permission to run CLUSTER NODES.
• Without UseCluster, only the connected endpoint is queried, even if it belongs to a Cluster. Cluster-only commands still fail on a standalone server.

var results = await redis.Server.ClusterShardsOnAllNodesAsync();
foreach (var result in results)
{
    if (result.IsSuccess)
        Console.WriteLine($"{result.Endpoint}: {result.Value.Length} reported shards");
    else
        Console.WriteLine($"{result.Endpoint}: {result.Error!.Message}");
}

• Check every result. Views can overlap or disagree during failover; nothing is merged.
• Cancellation during discovery throws. After discovery it is recorded per node, and completed results stay available.

These 20 methods are IRespireServerCommands members; external implementations, decorators, and mocks must implement or forward them. They are immediate, with no batch or transaction versions.