yanxiaole commented on pull request #29392:
URL: https://github.com/apache/spark/pull/29392#issuecomment-670961239
> What code that modifies this object is also synchronizing on this?
Sorry, I'll update them.
@srowen , do you think I should add synchronizing at every write and del
yanxiaole commented on pull request #29392:
URL: https://github.com/apache/spark/pull/29392#issuecomment-670993081
It happened when data structure getting modified, mainly `delete` caused the
problem.
Take `LevelDB` implementation for example, in its iterator's `next` function
it wi
yanxiaole commented on pull request #29392:
URL: https://github.com/apache/spark/pull/29392#issuecomment-671067561
Thanks @srowen , I just updated a new commit to lock the delete operation.
This is an automated message fr
yanxiaole commented on pull request #29392:
URL: https://github.com/apache/spark/pull/29392#issuecomment-671069843
I have pasted two stack traces below as they happened in two different
functions. The line numbers are slightly different with the code in `master`
because I'm using `branch-3
yanxiaole commented on pull request #29392:
URL: https://github.com/apache/spark/pull/29392#issuecomment-672459512
Hi @zhouyejoe ,
Yes, you are right. I have checked the pool definition, and `checkForLogs()`
and `cleanLogs()` won't run at same time, the only possible is the replay
tasks
yanxiaole commented on pull request #29392:
URL: https://github.com/apache/spark/pull/29392#issuecomment-672630663
the filter is called after the assignment of variable `stale`, but the race
condition is happened in the assignment, the `asScala.toList`.
---