[SCB-751] The load balancer of Saga may be faced with peak problem.

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: saga-0.2.0
• Fix Version/s: pack-0.7.0
• Component/s: Saga
• Labels: load-balancer saga
• Environment: project: incubator-servicecomb-saga-0.2.0
• External issue URL: https://github.com/apache/incubator-servicecomb-saga/issues/222
• Assignee: justin li
• Reporter: Gannalyo
• Created: 17/Jul/18 09:15
• Updated: 22/May/20 06:49

Description

Currently, Saga's load balancer will select a fastest 'MessageSender' to send msg from pool.

My problem:
At some point, if every Alpha node need take 10 minutes to handle business for some reasons(e.g. network), that means, Saga will handle all following requests by the fastest 'MessageSender' within 10 minutes.
So, I think that the performance for Saga's load balancer will get a discount.

Of course, I believe one alpha node can handle many many requests. Just for improving Saga.

Part of code from LoadBalancedClusterMessageSender.java  

... 
MessageSender messageSender = fastestSender(); 
try { 
    long startTime = System.nanoTime();
    AlphaResponse response = messageSender.send(event); 
    senders.put(messageSender, System.nanoTime() - startTime);
    return response; 
} catch (OmegaException e) {
 throw e; 
}

Issue Links

• relates to: SCB-745 RetrySender should not throw exception when go the saga start event.