Files
fk-wazuh/docs/ref/performance.md
T

2.6 KiB

Performance

This section provides practical recommendations to improve performance for Wazuh Docker deployments (single-node and multi-node). Apply the controls that match your workload and environment.

Performance drivers

  • Wazuh Indexer is typically the main bottleneck (JVM heap, disk I/O, and CPU).
  • Wazuh Manager load grows with the number of connected agents and event throughput.
  • Wazuh Dashboard mainly affects interactive usage and depends on Indexer responsiveness.

For baseline host sizing and prerequisites, see Requirements.

Storage and host

  • Use low-latency storage for the Indexer data volume (see Requirements).
  • Avoid slow or inconsistent storage for the Indexer (for example, network filesystems) unless you have validated latency and durability for your use case.
  • Monitor disk space growth. Index data and persistent volumes can grow quickly in high-ingest environments.

Wazuh Indexer (OpenSearch)

  • Set the JVM heap explicitly using OPENSEARCH_JAVA_OPTS (documented in Environment variables).
  • Keep heap sizing conservative relative to available memory so the OS can cache filesystem data; oversized heap commonly degrades disk-heavy workloads.
  • Ensure the Linux host meets the required vm.max_map_count prerequisite (documented in Requirements).
  • Prioritize heap sizing and GC stability.
  • Prioritize disk throughput/latency for the Indexer data volume.
  • Prioritize CPU availability during ingest peaks.

Wazuh Manager

  • If you observe ingestion backpressure or delayed processing, validate that the Manager has sufficient CPU and memory and that persistent volumes are not constrained by slow storage.
  • For multi-node deployments, distribute agent load appropriately (for example, by separating responsibilities between master/worker nodes) to avoid overloading.

Dashboard

  • Dashboard responsiveness depends on Indexer health. Address Indexer resource constraints first when troubleshooting slow UI queries.
  • Avoid exposing the Dashboard to excessive concurrent users on small hosts; scale the host or deployment model if needed.

Observability and troubleshooting

  • Start with container-level signals: docker stats, container logs, and host disk utilization.
  • For Indexer issues, validate basic cluster health and look for sustained CPU saturation, JVM memory pressure, and disk I/O contention.
  • For Manager issues, review Manager logs for queue growth and repeated connection retries.