← Back to the chapter

Technical detail · AMQP Reliability: Testing and Pitfalls

Adapt prefetch within a tested ceiling

This optional consumer example collects 100 processing times before adjusting its prefetch. The default target latency is 100 ms. The calculated reserve has a minimum of one and a ceiling of 20, so adaptation cannot create an unlimited window. Changing that window does not increase a sequential worker’s processing capacity. Test fairness with mixed consumer speeds before using it.

# OPTIONAL: Adaptive prefetch, with a configured ceiling and fairness test
class AdaptiveConsumer:
    def __init__(self, channel, target_latency_ms=100):
        self.channel = channel
        self.target_latency = target_latency_ms / 1000
        self.processing_times = []
        self.current_prefetch = 1

    def adjust_prefetch(self, processing_time):
        self.processing_times.append(processing_time)
        if len(self.processing_times) >= 100:
            avg_time = sum(self.processing_times) / len(self.processing_times)
            # This bounds the reserve; it does not increase sequential worker capacity.
            optimal = min(20, max(1, int(self.target_latency / avg_time * 2)))
            if optimal != self.current_prefetch:
                self.channel.basic_qos(prefetch_count=optimal)
                self.current_prefetch = optimal
            self.processing_times = []

← Back to the chapter