These are cluster redirects telling your client that the key lives on a different node. A cluster-aware client follows them automatically; a plain client surfaces them as errors. Jump to your situation below or work through the methods in order.
In a Redis Cluster, MOVED and ASK are redirects, not failures. The keyspace is split into 16384 hash slots spread across nodes, and when you send a command to a node that does not own the key's slot, it replies with the address of the node that does. MOVED means the slot has permanently moved there; ASK means a temporary redirect during a slot migration. A cluster-aware client follows these redirects automatically and caches the slot map, so your application never sees them. They appear as errors only when you use a plain single-node client (or redis-cli without -c) against a cluster. The fix is to use a cluster client (or pass -c to redis-cli), which makes MOVED and ASK invisible. Keeping the client's slot map fresh handles ongoing resharding.
A Redis Cluster shares 16384 hash slots across its nodes, and each key maps to one slot. If your command reaches a node that does not hold that slot, the node does not proxy it; instead it tells you where the slot lives with MOVED, or, mid-migration, with ASK.
These are normal cluster mechanics. A cluster-aware client expects them, follows the redirect, and remembers the slot map so the next request goes straight to the right node. You only see MOVED and ASK as errors when the client does not understand clustering and treats the redirect as a failure.
“MOVED and ASK are not errors, they are the cluster giving you directions. The keyspace is split into slots across nodes, and if you knock on the wrong node it says the key you want is over there. A cluster-aware client just follows that pointer and remembers it, so your app never notices. You only see these as errors when you point a plain client, or redis-cli without -c, at a cluster. Switch to a cluster client and they vanish.”
redis-cli -c -h host -p 6379
CLUSTER SLOTS.{user100}:profile.MOVED and ASK are cluster redirects, not failures: the key lives on another node. Use a cluster-aware client (or redis-cli with -c) so it follows them automatically and caches the slot map. Hash-tag related keys to co-locate them, and refresh the slot map after resharding.
If a cluster client still throws redirects, its node list may be incomplete or behind a NAT that rewrites the advertised addresses; checking CLUSTER NODES and the cluster-announce settings clarifies the topology the client sees. You can also submit your error to us for a tailored fix.