Elasticsearch High JVM Memory Pressure and Out of Memory

How to fix it

Try these in order — the most common fix is first.

1. 1. Check the right metric

   Copy steps
   1. Track old-generation usage as a rolling average, not instant heap percent
   2. Healthy GC: young under 50 ms about every 10 s, old under 1 s no more than once per 10 min
   Did this fix it? Yes No

2. 2. Fix heap and OS settings

   Copy steps
   1. Let Elasticsearch size the heap by default; if manual, Xms equals Xmx and at most half of RAM
   2. Disable swap on the OS
   3. Verify compressed OOPs are enabled
   Did this fix it? Yes No

3. 3. Cut memory consumers

   Copy steps
   1. Consolidate into fewer, larger shards
   2. Force merge or use an ILM force merge action after writes stop
   3. Lower index.max_result_window and search.max_buckets, set a default search timeout
   Did this fix it? Yes No

4. 4. Scale if the workload outgrew the hardware

   Copy steps
   1. Add node memory or nodes
   2. Enable AutoOps or stack monitoring alerts on JVM memory
   Did this fix it? Yes No

Symptoms

Nodes go slow, long GC pauses in logs, nodes drop out of the cluster, circuit breaker errors, occasionally OutOfMemoryError.

Why it happens

• Too many shards or segments
• Swap enabled
• Heap set above about 30 GB, which disables compressed OOPs
• Heap larger than half of system RAM
• Expensive aggregations, large result windows or bulk requests
• Mapping explosion

This guide also covers

• elasticsearch high jvm memory pressure
• elasticsearch out of memory
• elasticsearch heap size too high

The same problem in other apps

All slow, laggy or resource-hungry guides →

• Meilisearch Indexing Slow (Meilisearch)
• Typesense Out of Memory (OOM) and RAM Sizing (Typesense)
• Algolia Search Slow or High Latency (Algolia)

Other Elasticsearch errors

• Elasticsearch Cluster Health Yellow: Causes and Fixes
• Elasticsearch Cluster Health Red: How to Fix It
• Elasticsearch Unassigned Shards: Why and How to Fix
• Elasticsearch circuit_breaking_exception: Data Too Large
• Elasticsearch Disk Watermark: Index Read-Only (FORBIDDEN/12)
• Elasticsearch Limit of Total Fields [1000] Exceeded
• Why Is Elasticsearch So Slow? Fixing Slow Searches
• Elasticsearch Cannot Change Field Mapping: Reindex Guide