How to Remove All Keys in a Redis OSS Cluster

Redis OSS Cluster spreads keys across multiple master shards, so running FLUSHALL on a single node only clears that shard. To remove all keys across the cluster, you must connect to every master shard or use a cluster-aware client that fans out the command. This guide introduces Cluster concepts and key commands, lists Prerequisites and safety requirements, provides Step-by-step CLI instructions to discover masters and flush them, shows Cluster-aware client automation with common libraries, outlines Performance best practices for large datasets and safe batch deletes, and ends with Troubleshooting for common pitfalls.

Prerequisites

• Access to the Redis cluster with redis-cli or a cluster-aware client library.

• Authentication credentials with permission to run dangerous commands like FLUSHALL.

• Awareness that FLUSHALL is destructive and cannot be undone.

Step-by-Step Instructions

1. Discover all master shards

Run the following command on any cluster endpoint:

redis-cli -c -h <cluster-host> -p <cluster-port> CLUSTER SLOTS

This lists the hash slot ranges and the corresponding master shard IPs and ports.

2. Flush each master shard

For each master node returned from CLUSTER SLOTS, run:

redis-cli -h <shard-ip> -p <shard-port> FLUSHALL

• Repeat for all master shards.

• Include authentication if needed:

  redis-cli -h <shard-ip> -p <shard-port> -a <password> FLUSHALL

3. Use a cluster-aware client library

Cluster-aware clients (redis-py, Jedis, go-redis, etc.) can issue FLUSHALL across all master nodes automatically.

Python example (redis-py):

from redis.cluster import RedisCluster

startup_nodes = [
    {"host": "127.0.0.1", "port": 7000},
    {"host": "127.0.0.1", "port": 7001},
    {"host": "127.0.0.1", "port": 7002}
]

rc = RedisCluster(startup_nodes=startup_nodes, decode_responses=True)
rc.flushall()

Best Practices for Performance

• Schedule carefully: FLUSHALL is blocking and may impact performance. Run during off-peak hours in production.

• Delete subsets safely: For selective deletion (e.g., user:*), use SCAN with UNLINK in batches, or Redis Insight bulk deletion, instead of blocking commands.

• Large clusters: For huge datasets (tens of millions of keys), pipelined UNLINK with batching reduces blocking and fragmentation. See bulk key deletion best practices.

Troubleshooting Common Issues

Summary Table: Typical CLI Flow for Cluster Flush