Harvester 1.6.0: HTTP/2 Transition for Elastic Output – Unresolved Retry Backoff Behavior

Background

Harvester 1.6.0 introduced HTTP/2 support for the elastic output plugin, replacing the implicit HTTP/1.1 default. The configuration now requires protocol: http2 and increases the default max_retries to 5. While the retry mechanism uses an exponential backoff algorithm, the multiplier and jitter parameters are not exposed in the public configuration.

Goal and Constraints

The objective is to determine how the retry backoff behaves under HTTP/2 transport, particularly when intermittent connection failures occur. Constraints include the experimental nature of HTTP/2 support, lack of documented backoff tuning options, and potential compatibility issues with existing Elasticsearch clusters.

Questions

• What are the actual multiplier and jitter values used by Harvester’s HTTP/2 retry backoff algorithm in version 1.6.0?
• Is there a way to expose or override these backoff parameters through configuration or environment variables?
• How does the backoff behavior compare between HTTP/1.1 and HTTP/2 in terms of retry timing and overall connection pool exhaustion?