hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-14401) Stamp failed appends with sequenceid too.... Cleans up latches
Date Fri, 11 Sep 2015 00:22:46 GMT

    [ https://issues.apache.org/jira/browse/HBASE-14401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14739907#comment-14739907
] 

stack commented on HBASE-14401:
-------------------------------

Hmm.. No. Let me think more about it. I remove the latches to have the test run more free
but the idea was to manufacture what used to make us hang... my latches are being interrupted
by the log rolling which is messing up this technique. Let me see if I can figure another
means of getting the same ugly pause in there.

> Stamp failed appends with sequenceid too.... Cleans up latches
> --------------------------------------------------------------
>
>                 Key: HBASE-14401
>                 URL: https://issues.apache.org/jira/browse/HBASE-14401
>             Project: HBase
>          Issue Type: Sub-task
>          Components: test, wal
>            Reporter: stack
>            Assignee: stack
>             Fix For: 2.0.0, 1.2.0, 1.3.0
>
>         Attachments: 14401.txt
>
>
> Looking in test output I see we can sometimes get stuck waiting on sequenceid... The
parent issues redo of our semantic makes it so we encounter failed append more often around
damaged WAL.
> This patch makes it so we stamp sequenceid always, even if the append fails. This way
all sequenceids are accounted for but more important, the latch on sequenceid down in WALKey
will be cleared.. where before it was not being cleared (there is no global list of outstanding
WALKeys waiting on sequenceids so no way to clean them up... we don't need such a list if
we ALWAYS stamp the sequenceid).



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message