Axon应用事件并行处理配置及相关疑问咨询

Hey there! Let's break down your questions around parallel event processing and event replay in Axon Framework 4.0.3 clearly:

1. Will the provided configuration take effect just by adding it? #

Short answer: Yes, but you need to ensure the [name] placeholder matches your actual event processor's name. Here's the full breakdown:

• The settings you listed are specifically for Tracking Event Processors (TEP), which is why mode=tracking is mandatory—these won't work for Subscribing Processors.
• If you're using the default processor (your @EventHandler methods don't specify a processorName), replace [name] with default (Axon's default name for unassigned handlers).
• For custom processors (defined via the processorName parameter in @EventHandler), use that exact name in the config.
• As long as the processor name is correct and your event store supports tracking (Axon Server, JDBC, and other standard stores do), these settings will apply automatically—no extra Java code is needed unless you're building custom processor beans.

2. Do I need to configure each processor individually, or is there a global approach? #

You have both options to fit your use case:

• Global fallback via default processor: Configure the default processor to apply settings to all unassigned @EventHandler methods. Example:

  axon.eventhandling.processors.default.initialSegmentCount=4
  axon.eventhandling.processors.default.threadCount=2
  axon.eventhandling.processors.default.mode=tracking
  

  Custom processors won't inherit these defaults automatically in 4.0.3—you'll need to define separate config for them if you want different values.
• Per-processor configuration: For custom processors, define settings using their specific name. This is perfect if you need different parallelization levels for different event groups (e.g., a high-throughput processor for order events, a smaller one for audit logs).

3. Will deleting the processor entry in entry_token trigger full event replay? Can I replay specific events? #

• Deleting the entry_token record: Yes—removing a processor's token will force that TEP to replay all events from the start of the event store. The token tracks the last processed event position; without it, the processor has no way to resume, so it starts over.
• Replaying specific events: Absolutely, you don't have to replay everything. Here are practical ways to target specific events/ranges:
  • Configure initial tracking token: Use properties to start from a specific point:

    # Start from the latest event (no replay)
    axon.eventhandling.processors.[name].initialTrackingToken=latest
    # Start from a specific timestamp (ISO-8601 format)
    axon.eventhandling.processors.[name].initialTrackingToken=2024-01-01T00:00:00Z
    # Start from a specific event sequence number
    axon.eventhandling.processors.[name].initialTrackingToken=sequence:1000
    

  • Programmatic token reset: Inject the TrackingEventProcessor bean and reset its token to a custom position. For example:

    trackingEventProcessor.resetTokens(streamingToken -> 
        // Logic to set a token starting after a specific event ID, time, or sequence
    );
    

  • Filter events during replay: If you only want to process certain event types, add a condition in your @EventHandler method or use Axon's event routing rules to exclude unwanted events during replay.

内容的提问来源于stack exchange，提问作者Ali_Nass