hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sean Busbey (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-12235) Backport to 0.94: HBASE-9002 TestDistributedLogSplitting.testRecoverdEdits should test correct region
Date Sat, 11 Oct 2014 07:47:34 GMT

     [ https://issues.apache.org/jira/browse/HBASE-12235?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Sean Busbey updated HBASE-12235:
--------------------------------
    Description: 
I had TestDistributedLogSplitting.testRecoveredEdits fail on a internal test rig. It's very
infrequent.

The proximal failure looks like HBASE-9002, the assertion finds 0 edits instead of 1000.

Looking through the logs, the #regions printed is 1

{code}
line 236        LOG.info("#regions = " + regions.size());
{code}

And then in the makeHLog method I never see any log entry for the actual writing

{code}
line 569    for (int i = 0; i < n; i++) {
line 570      LOG.info("region " + hris.get(i).getRegionNameAsString() + " has " + counts[i]
+ " edits");
line 571    }
{code}

I think what's happening is we're picking a region that is META or ROOT, so the checks after
the "#regions=" line removes the region and we end up inserting no data.

It looks like the patch for HBASE-9002 should fix this.

  was:
I had TestDistributedLogSplitting.testRecoveredEdits fail on a internal test rig. It's very
infrequent.

The proximal failure looks like HBASE-9002, the assertion finds 0 edits instead of 1000.

Looking through the logs, the #regions printed is 0

{code}
line 236        LOG.info("#regions = " + regions.size());
{code}

And then in the makeHLog method I never see any log entry for the actual writing

{code}
line 569    for (int i = 0; i < n; i++) {
line 570      LOG.info("region " + hris.get(i).getRegionNameAsString() + " has " + counts[i]
+ " edits");
line 571    }
{code}

I think what's happening is we're picking a region that is META or ROOT, so the checks after
the "#regions=" line removes the region and we end up inserting no data.

It looks like the patch for HBASE-9002 should fix this.


> Backport to 0.94: HBASE-9002 TestDistributedLogSplitting.testRecoverdEdits should test
correct region
> -----------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-12235
>                 URL: https://issues.apache.org/jira/browse/HBASE-12235
>             Project: HBase
>          Issue Type: Task
>          Components: test
>    Affects Versions: 0.94.0
>         Environment: 0.94.15 compiled with jdk6 running on jdk7 (but I don't think it's
relevant)
>            Reporter: Sean Busbey
>            Assignee: Sean Busbey
>            Priority: Minor
>             Fix For: 0.94.25
>
>
> I had TestDistributedLogSplitting.testRecoveredEdits fail on a internal test rig. It's
very infrequent.
> The proximal failure looks like HBASE-9002, the assertion finds 0 edits instead of 1000.
> Looking through the logs, the #regions printed is 1
> {code}
> line 236        LOG.info("#regions = " + regions.size());
> {code}
> And then in the makeHLog method I never see any log entry for the actual writing
> {code}
> line 569    for (int i = 0; i < n; i++) {
> line 570      LOG.info("region " + hris.get(i).getRegionNameAsString() + " has " + counts[i]
+ " edits");
> line 571    }
> {code}
> I think what's happening is we're picking a region that is META or ROOT, so the checks
after the "#regions=" line removes the region and we end up inserting no data.
> It looks like the patch for HBASE-9002 should fix this.



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

Mime
View raw message