Troubleshooting High Memory Usage in Redis Cloud

Redis Cloud databases enforce a configured memory limit. When usage approaches that limit, Redis must evict keys or reject writes. This article provides step-by-step diagnostics, short-term mitigation strategies, and long-term architectural fixes to resolve sustained memory pressure. 

For warning interpretation and decision guidance, see High Memory Usage Warning in Redis Cloud: What It Means and What to Do Next.

Quick Diagnostic Checklist

Confirm utilization and trend

Redis Cloud Console → Database → Metrics
Review:

• Used memory vs limit

• Growth trend

• Evictions

• OOM errors

Optional CLI:

INFO memory

Focus on:

• used_memory

• maxmemory

• evicted_keys

• mem_fragmentation_ratio

Identify Large Keys

Sample large keys:

redis-cli --bigkeys

Large hashes, lists, sorted sets, streams, or JSON documents are common causes.

Unbounded collections are high-risk patterns.

Check Slow Commands

slowlog get 200

Look for:

• KEYS

• Full scans

• HGETALL or LRANGE 0 -1 on large structures

• JSON.GET on entire large documents

Memory pressure and latency frequently correlate.

Inspect Index Overhead

If using Search or JSON:

FT.INFO <index>

Review:

• num_docs

• Schema fields

• SORTABLE fields

• Unnecessary TEXT fields

Common issues:

• Indexing fields not queried

• Overuse of TEXT instead of TAG

• Missing NOINDEX for non-searchable fields

Indexes can consume significant RAM.

Short-Term Mitigation

Delete Unused Data

Use SCAN and UNLINK, not KEYS and DEL.

Example pattern-based cleanup:

--scan --pattern 'prefix:*'
UNLINK

Avoid blocking deletions in production.

Tighten TTL Policies

Ensure cache keys have expiration.
Refactor unbounded structures into time-bucketed keys with TTL.

Increase Database Size

If workload is legitimate and persistent, increase database size in the Redis Cloud Console.

Maintain headroom above peak expected usage.

Long-Term Fixes

Enforce Data Lifecycle

Define retention per data type:

• TTL

• Archival

• Cleanup jobs

Refactor Large Structures

Avoid:

• Infinite logs in lists or streams

• Ever-growing hashes

• Massive single JSON documents

Use:

• Time buckets

• Trimming

• Key sharding

Optimize Index Schemas

• Index only searchable fields

• Avoid unnecessary SORTABLE

• Use TAG for categorical fields

• Scope indexes using PREFIX

Rebuild carefully and measure before and after.

Scale Horizontally

If growth is real and sustained:

• Increase memory

• Enable clustering

• Ensure cluster-aware clients

Monitoring Best Practices

Track:

• used_memory vs maxmemory

• evicted_keys

• OOM errors

• Latency percentiles

Recommended alerts:

• Warning at 75 to 80 percent

• Critical at 90 percent

• Immediate alert on OOM

Summary

High memory pressure in Redis Cloud is typically caused by one or more of:

• Unbounded data growth

• Missing TTL policies

• Oversized keys

• Over-indexed datasets

• Legitimate, sustained growth without scaling

Durable resolution requires lifecycle enforcement, schema optimization, and right-sizing capacity.