[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17175126#comment-17175126 ] Hudson commented on HBASE-23157: Results for branch branch-2.2 [build #931 on

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17174958#comment-17174958 ] Hudson commented on HBASE-23157: Results for branch master [build #1809 on

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17174445#comment-17174445 ] Hudson commented on HBASE-23157: Results for branch branch-2.3 [build #205 on

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16960614#comment-16960614 ] Duo Zhang commented on HBASE-23157: --- Reimplement the UT so we do not need to add a dummy method for

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16957830#comment-16957830 ] Lijin Bin commented on HBASE-23157: --- TestMaxFlushedSeqId.patch can reproduce the problem that a last

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16955691#comment-16955691 ] Lijin Bin commented on HBASE-23157: --- Yes, we must make sure that before calling startCacheFlush we

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16955688#comment-16955688 ] Duo Zhang commented on HBASE-23157: --- OK, so the problem is that, in startCacheFlush we will remove

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16955685#comment-16955685 ] Lijin Bin commented on HBASE-23157: --- [~zhangduo] The wal.sync is called after

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16955490#comment-16955490 ] Duo Zhang commented on HBASE-23157: --- [~binlijin] After checking the code, we do have a wal.sync when

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16951854#comment-16951854 ] Lijin Bin commented on HBASE-23157: --- I think the first choice is to force a wal.sync before

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16950868#comment-16950868 ] Duo Zhang commented on HBASE-23157: --- OK, so this is the problem. Let me think how to fix it. > WAL

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16950866#comment-16950866 ] Guangxu Cheng commented on HBASE-23157: --- HI, [~stack] , [~zhangduo] this issue does not cause data

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16950717#comment-16950717 ] Lijin Bin commented on HBASE-23157: --- [~zhangduo] this may cause regionserver have too many wals and

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16950716#comment-16950716 ] Lijin Bin commented on HBASE-23157: --- [~stack] We find a cluster with some regionserver have 20

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16950242#comment-16950242 ] Duo Zhang commented on HBASE-23157: --- So the problem here is that the unflushedSequenceId go backward

[jira] [Commented] (HBASE-23157) WAL unflushed seqId tracking may wrong when Durability.ASYNC_WAL is used

[ https://issues.apache.org/jira/browse/HBASE-23157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16950082#comment-16950082 ] Michael Stack commented on HBASE-23157: --- How did you figure this one out [~binlijin]? You are