CLUSTERDOWN error

The cluster can't serve the command because a hash slot has no working node.

A Redis cluster returns this error when it can't serve your command. The message has two versions:

CLUSTERDOWN Hash slot not served
CLUSTERDOWN The cluster is down

• Hash slot not served means that no node serves the key's hash slot. This happens when some slots were never assigned to a node, for example, in a cluster you haven't finished setting up.
• The cluster is down means that the cluster is in a failed state. By default, a cluster stops serving commands when any hash slot has no working node. For example, this happens when a primary node fails and has no replica to take its place.

In redis-py, this error is ClusterDownError: The cluster is down or ClusterDownError: Hash slot not served.

Example

After a primary with no replica fails, the other nodes stop serving commands, even for their own keys:

> GET foo
(error) CLUSTERDOWN The cluster is down

How to fix it

1. Check the state of the cluster with CLUSTER INFO. cluster_state:fail means the cluster is down. cluster_slots_assigned shows how many of the 16384 slots have a node.
2. Find the failed nodes with CLUSTER NODES. A failed node has fail in its flags.
3. Bring the failed node back, or move its slots to working nodes. If slots were never assigned, assign them. redis-cli --cluster create assigns all the slots when it creates a cluster.

To keep the cluster working when a primary fails, give each primary at least one replica. A replica then takes its primary's place.

Two settings change what a cluster does when some slots have no working node:

• cluster-require-full-coverage no lets nodes keep serving the slots that still have a working node.
• cluster-allow-reads-when-down yes lets nodes serve reads while the cluster is in a failed state. Writes then get CLUSTERDOWN The cluster is down and only accepts read commands.

See Redis Cluster configuration parameters.