backpressure.md

\page backpressure Queue Back-Pressure Controls

Queue Back-Pressure Controls

The global asynchronous task executor backs loggers that use the default executor and can be tuned to handle high-load bursts. Some backends are synchronous, while configured dedicated executors and OTLP maintain their own queues. Use the following helpers from <logit.hpp> when preparing stress tests or long-running services:

• LOGIT_SET_MAX_QUEUE(size) sets the maximum number of queued tasks. Use a small size to emulate a constrained environment or 0 to remove the bound completely.
• LOGIT_SET_QUEUE_POLICY(mode) switches the overflow strategy. Available options are LOGIT_QUEUE_BLOCK, LOGIT_QUEUE_DROP_NEWEST, and LOGIT_QUEUE_DROP_OLDEST.
• LOGIT_GET_DROPPED_TASKS() returns the number of tasks that were discarded under the current configuration. The value is safe to read concurrently with producers.
• LOGIT_RESET_DROPPED_TASKS() clears the drop counter. Call it between scenarios so each run measures its own losses.

The drop counter is maintained inside the executor and is updated every time a publishing policy decides to discard work. Combining the counter with TaskExecutor::wait() makes it easy to assert the expected throughput for each policy without inspecting private state.

These global controls affect the shared TaskExecutor only. A backend with Config::use_dedicated_executor=true has its own capacity and policy, and OTLP uses its exporter queue settings. After finishing a burst test remember to restore the shared capacity or shut down the logging subsystem with LOGIT_WAIT() and LOGIT_SHUTDOWN() to avoid interfering with other scenarios.