[GitHub] [hadoop-ozone] runzhiwang commented on pull request #1028: HDDS-3735. Improve SCM performance with 3.7% by remove unnecessary lock and unlock

2020-06-09 Thread GitBox
runzhiwang commented on pull request #1028: URL: https://github.com/apache/hadoop-ozone/pull/1028#issuecomment-641678745 @elek Could you help review this patch ? Thank you very much. This is an automated message from the Apac

[GitHub] [hadoop-ozone] runzhiwang commented on pull request #1028: HDDS-3735. Improve SCM performance with 3.7% by remove unnecessary lock and unlock

2020-06-22 Thread GitBox
runzhiwang commented on pull request #1028: URL: https://github.com/apache/hadoop-ozone/pull/1028#issuecomment-647830265 @elek Thanks for review. You are right, if the write code like following, maybe after `currentInfo.setState(newState)` and before `currentInfo.setOwner(owner)`, we read

[GitHub] [hadoop-ozone] runzhiwang commented on pull request #1028: HDDS-3735. Improve SCM performance with 3.7% by remove unnecessary lock and unlock

2020-06-30 Thread GitBox
runzhiwang commented on pull request #1028: URL: https://github.com/apache/hadoop-ozone/pull/1028#issuecomment-651703240 > Without write lock the container might be removed from ownerMap and not from the typeMap, which can cause inconsistency. I think this case can not be avoided by