bugfix: fix not return value when branchCommit and branchRollback throw exception #2755
Conversation
Codecov Report
@@ Coverage Diff @@
## develop #2755 +/- ##
=============================================
+ Coverage 50.35% 50.49% +0.13%
Complexity 2919 2919
=============================================
Files 572 572
Lines 18612 18613 +1
Branches 2212 2242 +30
=============================================
+ Hits 9373 9398 +25
+ Misses 8286 8285 -1
+ Partials 953 930 -23 |
| @@ -52,7 +52,7 @@ | |||
| out.writeShort((short)0); | |||
| } | |||
| out.writeLong(branchId); | |||
| out.writeByte(branchStatus.getCode()); | |||
| out.writeByte((branchStatus == null ? BranchStatus.Unknown : branchStatus).getCode()); | |||
There was a problem hiding this comment.
I think the main problem is that what kind of situation would let the branch status null in branch commit or branch rollback.
There was a problem hiding this comment.
because when branch commit or rollback, the branch status should not be null.
There was a problem hiding this comment.
OK, I'll find out why.
There was a problem hiding this comment.
I think the main problem is that what kind of situation would let the branch status null in branch commit or branch rollback.
I found the cause of the problem and fixed it.
wangliang181230
changed the title
|
does it break consistency without retrying? |
嗯，我也考虑过。但是重试的频率太快了，如果同时有很多个事务都这样在重试，日志都要爆了。 |
我觉得你可以换个思路,给重试加上上限值,这个上限值可以通过配置中心获取.超过上限值后日志输出一次需要人工介入处理的日志. |
嗯，不断重试的原因肯定是由系统来找原因。只是我觉得seata最好也能提供一个重试的规则配置。 |
我觉得你可以把这个pr转化为这个方向,加入一个重试的规则配置.原因的话seata一般有打印,比如脏读写的镜像数据与当前匹配不上. |
wangliang181230
changed the title
wangliang181230
changed the title
There was a problem hiding this comment.
Rollback fail retrying we should talk it in issue. Let us solve the problem that make the branch status null first. I already check the logic in
io.seata.rm.datasource.DataSourceManager#branchRollback
io.seata.rm.datasource.DataSourceManager#branchCommit
But look like the branch status would not be null.
Can you check these classes make the branch status null?
SagaResourceManager
TCCResourceManager
ResourceManagerXA
DataSourceManager
RM和分支事务的状态并没有问题，而是出现异常时没有将状态写入BranchRollbackResponse中，从而导致AbstractBranchEndResponseCodec输出响应时，抛了空指针异常。 |
2）TCC的RM的二阶段不再往外抛异常，而是直接“return BranchStatus.PhaseTwo_XxxxxxFailed_Retryable;”.
wangliang181230
changed the title
|
现在换了一种处理方式，不是由AbsRMHandler去补response的数据，而是由各RM拦截异常并返回相应的BranchStatus，来避免因异常导致response数据未写入。
…------------------ 原始邮件 ------------------
发件人: "FUNKYE"<notifications@github.com>;
发送时间: 2020年6月5日(星期五) 下午5:08
收件人: "seata/seata"<seata@noreply.github.com>;
抄送: "王良"<841369634@qq.com>;"State change"<state_change@noreply.github.com>;
主题: Re: [seata/seata] bugfix: Response data not written when commit or rollback failed in tcc resource manager. (#2755)
@a364176773 approved this pull request.
LGTM
In rm/src/main/java/io/seata/rm/AbstractRMHandler.java:
> response.setXid(xid); response.setBranchId(branchId); - response.setBranchStatus(status); - if (LOGGER.isInfoEnabled()) { - LOGGER.info("Branch commit result: " + status); + try {
这块代码改动的目的是?原来是直接拿到status放入即可.doBranchCommit ,doBranchRollback都一样
—
You are receiving this because you modified the open/close state.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
slievrly
changed the title
Ⅰ. Describe what this PR did
bugfix: TCC分支事务，在commit或rollback出现异常时，response不会写入分支事务数据的BUG修复。
此BUG会造成输出响应时抛出空指针异常。代码行：
io.seata.serializer.seata.protocol.transaction.AbstractBranchEndResponseCodec中的
55行 out.writeByte(branchStatus.getCode());Ⅱ. Does this pull request fix one issue?
Ⅲ. Why don't you add test cases (unit test/integration test)?
Ⅳ. Describe how to verify it
Ⅴ. Special notes for reviews
目前还遗留了一直异常时不断且快速重试的问题。需要新的重试机制的PR来解决此问题。