ActiveMQ Message Queue Optimization: Removing Synchronized Lock and Tuning queuePrefetch

Overview: The production environment experienced severe message‑queue backlog in ActiveMQ, causing transaction delays.

Problem analysis: High volume of notification messages during traffic spikes, and the onMessage listener was synchronized, limiting concurrency to a single thread.

Code snippet showing the synchronized onMessage method:

public synchronized void onMessage(Message message, Session session)

Removing the synchronized lock allowed concurrent processing, reducing processing time from 151 s to 13 s for 15 000 messages.

Further tuning: Adjusted the queuePrefetch parameter from the default 1000 to 100, which increased the number of active consumers and reduced total processing time to 6 s.

Configuration example:

tcp://localhost:61616?jms.prefetchPolicy.queuePrefetch=100

Second‑stage improvement: Replaced a single delay queue with a dual‑queue design to isolate failed notifications, preventing them from blocking successful ones.

Third‑stage suggestion: Consider replacing ActiveMQ with higher‑throughput brokers such as RabbitMQ, RocketMQ or Kafka.

Conclusion: By removing the synchronized lock, tuning queuePrefetch, and adopting a dual‑queue strategy, the message‑consumption capacity increased more than 30‑fold, effectively solving the backlog issue.