How to fix MOVED Hash slot served by different node in Redis

The simplest solution is to use a Redis client that natively supports cluster mode. Most popular Redis clients (redis-py, node-redis, Lettuce, Jedis) handle MOVED redirects automatically.

For Python (redis-py):

from redis.cluster import RedisCluster

# Cluster-aware client automatically handles redirects
rc = RedisCluster(startup_nodes=[{"host": "localhost", "port": 7000}])
rc.set("mykey", "myvalue")

For Node.js (node-redis):

const { createCluster } = require("redis");

const cluster = createCluster({
  rootNodes: [
    { host: "localhost", port: 7000 },
  ],
});

await cluster.connect();
await cluster.set("mykey", "myvalue");

Cluster-aware clients automatically:
- Discover all nodes in the cluster
- Map keys to their hash slots
- Route requests to the correct node
- Handle MOVED and ASK redirects transparently

If your client doesn't automatically handle redirects, refresh the cluster topology when you receive a MOVED error.

The client should execute the CLUSTER SHARDS (or deprecated CLUSTER SLOTS) command to get the latest mapping:

redis-cli -c CLUSTER SHARDS

This returns all nodes and their slot ranges. Your client should:
1. Cache this topology
2. Update it whenever a MOVED error is received
3. Use the new mapping for future commands

Many clients support this with a configuration option like auto_refresh_on_moved or similar.

If you're testing with redis-benchmark against a cluster, use the -c flag to enable cluster mode:

redis-benchmark -h 127.0.0.1 -p 7000 -c -q -n 100000

The -c flag makes redis-benchmark cluster-aware, so it can follow MOVED redirects and correctly distribute keys across nodes.

Without -c, redis-benchmark sends all commands to a single node, triggering MOVED errors for keys that belong to other nodes.

Check that your Redis Cluster is healthy and all slots are assigned:

redis-cli -c CLUSTER INFO
redis-cli -c CLUSTER NODES

Look for:
- "cluster_state:ok" (not "fail")
- All 16384 slots assigned to nodes
- No nodes in "migrating" or "importing" state (unless actively resharding)

If you see "cluster_slots_fail" or unassigned slots, the cluster topology is broken. Run:

redis-cli --cluster fix <any-node>:6379

to attempt automatic repair.

If implementing a custom Redis client, handle MOVED errors by parsing the response and retrying:

// Pseudo-code
function executeCommand(command, key) {
  const slot = crc16(key) % 16384;
  const node = topologyMap[slot];

  try {
    return send(node, command);
  } catch (err) {
    if (err.message.includes("MOVED")) {
      // Parse: "MOVED <slot> <host>:<port>"
      const [_, newSlot, hostPort] = err.message.split(" ");
      const [host, port] = hostPort.split(":");

      // Update topology
      topologyMap[newSlot] = { host, port };

      // Retry command against new node
      return send({ host, port }, command);
    }
    throw err;
  }
}

Additionally, periodically refresh the cluster topology to proactively catch topology changes.