QdrantKubernetes

Page Faults Indicate Memory Swapping

critical
Resource ContentionUpdated Feb 23, 2026

Major page faults (process_major_page_faults_total) increase, indicating the OS is swapping vector data to disk and severely degrading query performance.

How to detect:

Alert on sustained increases in process_major_page_faults_total. Cross-reference with container_memory_working_set_bytes approaching limits and query latency spikes in rest_responses_duration_seconds.

Recommended action:

Immediately increase memory limits to eliminate swapping. Enable on_disk storage for vectors (qdrant_collection_config_params_always_ram=false) or implement quantization to reduce resident memory requirements.