Smart Mirroring Clone Latency Under Concurrent Requests

This piece investigates the latency behavior of the Smart Mirroring feature in Bitbucket Server when handling concurrent read‑only clone requests from multiple clients. The goal is to determine whether observed response delays stem from mirror instance thread allocation, database pool contention, or an undocumented request‑queuing policy.

Key uncertainties include how the mirror instance prioritizes incoming clone operations when its internal thread pool is saturated, whether the primary server’s server.maxThreads setting influences mirror request acceptance, and if any configuration exists to throttle or queue concurrent clones without disabling the feature. Current documentation does not specify a back‑pressure mechanism for mirror‑side event handling.

• How does the Smart Mirroring architecture prioritize concurrent clone requests when the mirror instance’s thread pool is saturated?
• What configuration parameters, if any, govern the mirror’s request acceptance threshold during high concurrency?
• Is there a documented mechanism for back‑pressure or queue management within the Smart Mirroring feature?