Mailing-List: contact issues-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Date: Wed, 17 Feb 2016 04:25:18 +0000 (UTC)
From: "Enis Soztutar (JIRA)" <jira@apache.org>
To: issues@hbase.apache.org
Message-ID: <JIRA.12938731.1455260825000.67775.1455683118193@Atlassian.JIRA>
In-Reply-To: <JIRA.12938731.1455260825000@Atlassian.JIRA>
References: <JIRA.12938731.1455260825000@Atlassian.JIRA>
 <JIRA.12938731.1455260825420@arcas>
Subject: [jira] [Updated] (HBASE-15259) WALEdits under replay will also be
 replicated
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/HBASE-15259?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Enis Soztutar updated HBASE-15259:
----------------------------------
    Status: Patch Available  (was: Open)

> WALEdits under replay will also be replicated
> ---------------------------------------------
>
>                 Key: HBASE-15259
>                 URL: https://issues.apache.org/jira/browse/HBASE-15259
>             Project: HBase
>          Issue Type: Bug
>            Reporter: ramkrishna.s.vasudevan
>            Assignee: ramkrishna.s.vasudevan
>            Priority: Minor
>         Attachments: HBASE-15259.patch
>
>
> I need to verify this but seeing the code
> {code}
> try {
>         // We are about to append this edit; update the region-scoped sequence number.  Do it
>         // here inside this single appending/writing thread.  Events are ordered on the ringbuffer
>         // so region sequenceids will also be in order.
>         regionSequenceId = entry.stampRegionSequenceId();
>         // Edits are empty, there is nothing to append.  Maybe empty when we are looking for a
>         // region sequence id only, a region edit/sequence id that is not associated with an actual
>         // edit. It has to go through all the rigmarole to be sure we have the right ordering.
>         if (entry.getEdit().isEmpty()) {
>           return;
>         }
>         // Coprocessor hook.
>         if (!coprocessorHost.preWALWrite(entry.getHRegionInfo(), entry.getKey(),
>             entry.getEdit())) {
>           if (entry.getEdit().isReplay()) {
>             // Set replication scope null so that this won't be replicated
>             entry.getKey().setScopes(null);
>           }
>         }
>         if (!listeners.isEmpty()) {
>           for (WALActionsListener i: listeners) {
>             // TODO: Why does listener take a table description and CPs take a regioninfo?  Fix.
>             i.visitLogEntryBeforeWrite(entry.getHTableDescriptor(), entry.getKey(),
>               entry.getEdit());
>           }
>         }
> {code}
> When a WALEdit is in replay we set the Logkey to null. But in the visitLogEntryBeforeWrite() we again set the LogKey based on the replication scope associated with the cells. So the previous step of setting null does not work here?


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