Massive Key Deletion in Redis Without Impacting Performance

Deleting a large number of keys in Redis can cause latency spikes or blocked clients if handled with synchronous commands like DEL or KEYS. This guide explains how to perform large-scale deletions safely across Redis Software, Redis Cloud, and Redis OSS without affecting performance. It includes guidance on Environment-Specific Behavior, explains Why Massive Key Deletion Impacts Performance, outlines Recommended Methods for Bulk Deletion using non-blocking operations like UNLINK and SCAN, provides detailed Step-by-Step Instructions for both CLI and Redis Insight, and concludes with Troubleshooting and Best Practices to ensure efficient cleanup and resource recovery in production environments.

Environment-Specific Behavior

Why Massive Key Deletion Impacts Performance

DEL is synchronous and blocks the main thread, so large deletions can cause latency spikes and stalled clients.
KEYS scans the full keyspace and should never be used in production.

Multi-key delete behavior depends on database type. Standard Redis Software clustered databases allow multi-key DEL or UNLINK when keys share a hash slot.
Active-Active CRDBs do not, and return:

CROSSSLOT Keys in request don't hash to the same slot

See clustering behavior for multi-key operations: Redis Clustering Overview

Recommended Methods for Bulk Deletion

1. Use the UNLINK Command (Recommended)
UNLINK is an asynchronous alternative to DEL.
It removes keys immediately from the keyspace and reclaims memory in the background.
Always use UNLINK for bulk deletions in production environments.

2. Use SCAN to Find Keys by Pattern
SCAN is a cursor-based, non-blocking iterator for safely listing keys.
Combine SCAN with UNLINK to delete keys matching a pattern.
Avoid using KEYS in production for this purpose.

3. Use Redis Insight Bulk Actions
Redis Insight provides a Bulk Actions tool that deletes matching keys using non-blocking operations.
Recommended for interactive cleanup or validation before deletion.

Step-by-Step Instructions

A. Command-Line (redis-cli) Method

Delete keys matching a pattern safely:

redis-cli -h <HOST> -p <PORT> --scan --pattern 'your:pattern:*' -i 0.01 | xargs redis-cli -h <HOST> -p <PORT> unlink

Explanation:

• <HOST> and <PORT> are the endpoint of your database.
• --scan --pattern safely iterates through keyspace.
• -i 0.01 adds a 10 ms delay between operations.
• xargs batches deletions; use -L <number> to limit keys per UNLINK call.

CRDB-Safe Method (Avoid CROSSSLOT Errors):

redis-cli -h <HOST> -p <PORT> --scan --pattern <PATTERN> -i 0.01 \
  | sed -e 's/\\/\\\\/g' -e 's/"/\\"/g' -e 's/\t/\\t/g' \
        -e 's/^/UNLINK "/' -e 's/$/"/' \
  | redis-cli -h <HOST> -p <PORT>

This method writes all matching keys to a file, prefixes them with UNLINK, and executes them one by one to ensure cross-slot safety.

Python Example (redis-py, RedisCluster):

from redis.cluster import RedisCluster

def delete_keys_by_pattern(pattern="user:*", scan_count=1000, pipe_batch=1000):
    cursor = "0"
    while True:
        cursor, keys = rc.scan(cursor=cursor, match=pattern, count=scan_count)
        if keys:
            for i in range(0, len(keys), pipe_batch):
                batch = keys[i:i+pipe_batch]
                pipe = rc.pipeline()
                for k in batch:
                    pipe.unlink(k)
                pipe.execute()
        if cursor == "0":
            break

This pattern-based deletion works across Redis Software standalone, clustered, and CRDB databases when the OSS API is enabled.

B. Redis Insight Bulk Actions

Redis Insight bulk deletion works on both standalone and CRDB databases.

1. Open Redis Insight and connect to your database.
2. Navigate to Browser → Tree/List View.
3. Filter by key name or type.
4. Open Bulk Actions → Delete Keys.
5. Review the list before deletion.
6. Click Delete to execute safely.

Quick Fix Table

Troubleshooting Common Issues

Best Practices

• Always prefer UNLINK over DEL for large deletions.
• Never use KEYS in production.
• Run deletions per shard or node in clustered setups.
• For CRDB, delete keys individually or with per-key scripts.
• Monitor latency, ops/sec, and memory usage during deletions.
• Use MEMORY PURGE after large deletes to reclaim space.
• Schedule major deletions during low-traffic or maintenance windows.
• Use Redis Insight Bulk Actions for visual safety and confirmation.

Troubleshooting Checklist

• Confirm you’re not using KEYS or DEL on large keyspaces.
• Validate ACL permissions for deletion commands.
• If memory isn’t freed, check for fragmentation or CRDB tombstones.
• In CRDB, ensure all deleted keys belong to the same hash slot or use per-key deletion.
• Use Redis Insight Profiler, SLOWLOG, or Prometheus to monitor during deletion.