Bulk Indexing Saturation

elasticsearch.index.operations.time for indexing increasing significantly with high elasticsearch.os.cpu.usage (>70% sustained) and growing elasticsearch.node.operations.current for index operations

Increase index.refresh_interval to 30s to reduce segment creation frequency (check current via elasticsearch.indices.settings). Optimize bulk request size - typically 5-15MB per request or 1000-5000 documents. Consider separating indexing and search workloads to dedicated nodes using shard allocation awareness. Add nodes to distribute indexing load if CPU-saturated. Monitor elasticsearch.node.thread_pool.tasks.queued for bulk/write thread pool saturation. Disable replicas temporarily during large bulk loads (_settings API), then re-enable after completion. Check elasticsearch.indexing_pressure.memory.limit and elasticsearch.indexing_pressure.memory.total.replica_rejections for memory-based throttling.