[GitHub] [hbase] sunhelly commented on pull request #2699: HBASE-25287 Forgetting to unbuffer streams results in many CLOSE_WAIT…

sunhelly commented on pull request #2699: URL: https://github.com/apache/hbase/pull/2699#issuecomment-741553118 Failed UT is not relevant to this issue. @Apache9 Could you help to review this? I have fixed this issue and deployed on our production clusters, all work well.

[GitHub] [hbase] sunhelly commented on pull request #2699: HBASE-25287 Forgetting to unbuffer streams results in many CLOSE_WAIT…

sunhelly commented on pull request #2699: URL: https://github.com/apache/hbase/pull/2699#issuecomment-734230527 TestRSGroupsFallback.testFallback always passes on my local computer. A fixed issue is proposed at `HBASE-25334 TestRSGroupsFallback.testFallback is flaky`

[GitHub] [hbase] sunhelly commented on pull request #2699: HBASE-25287 Forgetting to unbuffer streams results in many CLOSE_WAIT…

sunhelly commented on pull request #2699: URL: https://github.com/apache/hbase/pull/2699#issuecomment-734046820 @Apache9 Yes, it's right. `initMetaAndIndex` is also used somewhere else. But I think `initTrailerAndContext` also needs to add a finally. PR has been updated.

[GitHub] [hbase] sunhelly commented on pull request #2699: HBASE-25287 Forgetting to unbuffer streams results in many CLOSE_WAIT…

sunhelly commented on pull request #2699: URL: https://github.com/apache/hbase/pull/2699#issuecomment-733530907 > Looks like it just adding a try/finally so we always unbuffer. Thats good. We weren't calling unbuffer ever? Thanks. Otherwise patch looks good to me. Thanks for

[GitHub] [hbase] sunhelly commented on pull request #2699: HBASE-25287 Forgetting to unbuffer streams results in many CLOSE_WAIT…

sunhelly commented on pull request #2699: URL: https://github.com/apache/hbase/pull/2699#issuecomment-733527278 > I see that in HFile.createReader we will call context.getInputStreamWrapper().unbuffer()? It is not enough to solve the problem here? Mind explaining a bit more? Thanks