v.25.4Improvement
Enabled a backoff logic
Enabled a backoff logic for all types of replicated tasks. It will provide the ability to reduce CPU usage, memory usage, and log file sizes. Added new settingsmax_postpone_time_for_failed_replicated_fetches_ms,max_postpone_time_for_failed_replicated_merges_msandmax_postpone_time_for_failed_replicated_tasks_mswhich are similar tomax_postpone_time_for_failed_mutations_ms. #74576 (MikhailBurdukov).
Why it matters
This feature addresses the problem of excessive CPU, memory, and log file consumption caused by immediate retries of failed replicated tasks. By introducing backoff delays, it reduces system load and improves cluster stability and performance.How to use it
Users can configure the backoff behavior using the new settings:max_postpone_time_for_failed_replicated_fetches_ms, max_postpone_time_for_failed_replicated_merges_ms, and max_postpone_time_for_failed_replicated_tasks_ms. These settings define the maximum postponement time (in milliseconds) for failed replicated fetches, merges, and other replicated tasks respectively, similar to the existing max_postpone_time_for_failed_mutations_ms setting.