bugfix: tcc phase two response timeout exception(#3484)

[zch0214]

Ⅰ. Describe what this PR did

TCC模式下，global commit的响应依赖于branch commit的响应。而global commit与branch commit的处理线程用的同一线程池，当并发量较大时，若线程池里都是global commit的线程，此时无法创建新线程执行branch commit的响应处理，导致超时。

Ⅱ. Does this pull request fix one issue?

fixes #3484
fixes #3711

Ⅲ. Why don't you add test cases (unit test/integration test)?

Ⅳ. Describe how to verify it

Ⅴ. Special notes for reviews

[xyz327]

我本地将seata服务核心线程池(ServerHanlderThread)设置为1 然后再使用 TCC 模式就会出现超时问题
这个是我复现的项目demo 以及问题分析 seata 版本 1.4.2 https://github.com/xyz327/seata-tcc-timeout-demo

[xyz327]

如果是单线程必然出现，那默认的50个线程在高并发情况下也会出现这个问题

[codecov-commenter]

Codecov Report

Merging #3497 (eae37b6) into develop (c60cb5e) will decrease coverage by 0.03%.
The diff coverage is 0.00%.

@@              Coverage Diff              @@
##             develop    #3497      +/-   ##
=============================================
- Coverage      49.43%   49.40%   -0.04%     
+ Complexity      3761     3759       -2     
=============================================
  Files            700      700              
  Lines          23626    23637      +11     
  Branches        2918     2918              
=============================================
- Hits           11679    11677       -2     
- Misses         10756    10767      +11     
- Partials        1191     1193       +2     

Impacted Files	Coverage Δ
...ava/io/seata/core/constants/ConfigurationKeys.java	0.00% <ø> (ø)
...a/io/seata/core/rpc/netty/NettyRemotingServer.java	0.00% <0.00%> (ø)
...ava/io/seata/core/rpc/netty/NettyServerConfig.java	0.00% <0.00%> (ø)
...o/seata/server/coordinator/DefaultCoordinator.java	47.84% <0.00%> (-0.54%)	⬇️
...in/java/io/seata/server/session/GlobalSession.java	79.56% <0.00%> (-0.44%)	⬇️

[Pinocchio2018]

请问，这个优化会被合并吗？我想了解下这个超时问题的最新进展。

will this change be merged? I want to know the latest progress on this problem.

[Pinocchio2018]

我想提个问题

public class RmBranchCommitProcessor implements RemotingProcessor {

    private void handleBranchCommit(RpcMessage request, String serverAddress, BranchCommitRequest branchCommitRequest) {
        BranchCommitResponse resultMessage = (BranchCommitResponse)this.handler.onRequest(branchCommitRequest, (RpcContext)null);
        if (LOGGER.isDebugEnabled()) {
            LOGGER.debug("branch commit result:" + resultMessage);
        }

        try {
            this.remotingClient.sendAsyncResponse(serverAddress, request, resultMessage);
        } catch (Throwable var6) {
            LOGGER.error("branch commit error: {}", var6.getMessage(), var6);
        }

    }
}

请问，this.remotingClient.sendAsyncResponse(serverAddress, request, resultMessage);这行代码，是在向tc发送二阶段成功的响应吗？这里是在向tc发送一个新的request吗？

如果是在向tc发req的话，我不理解，为什么rm收到tc的调用后，不是return结果，而是新发送一个req给tc呢？

[l81893521]

LGTM

[l81893521]

This change is important, please @slievrly review it.

[l81893521]

We talk about your solution lately, and we agree about that. Can you change the 1.5.0.md?

[zch0214]

We talk about your solution lately, and we agree about that. Can you change the 1.5.0.md?

Sure.Thanks for merging.

[funky-eyes]

LGTM