dfs datanode metrics, bytes_read, bytes_written overflows due to incorrect type used.

• Key: HADOOP-4228
• Type: Bug
• Status: Closed
• Resolution: Fixed
• Priority: Blocker
• Assignee: Eric Yang
• Reporter: Eric Yang
• Component/s: metrics
• Affects Version/s: 0.18.2, 0.19.0
• Fix Version/s: 0.18.2

Description

bytes_read, and bytes_written metrics are using int (MetricsTimeVaryingInt) as counter. This type is too small to store the bytes_read and bytes_written metrics. Recommend to change this to long (metricsLongValue).

Comments

[Eric Yang]

Patch to change metrics type from Integer to Long.

[Eric Yang]

Patch to change bytes_read and bytes_written metrics from type Integer to type long.

[Eric Yang]

Same patch for trunk.

[Hadoop QA]

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12390702/HADOOP-4228-trunk.patch
against trunk revision 697306.

+1 @author. The patch does not contain any @author tags.

+1 tests included. The patch appears to include 6 new or modified tests.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

-1 core tests. The patch failed core unit tests.

-1 contrib tests. The patch failed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3349/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3349/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3349/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3349/console

This message is automatically generated.

[Eric Yang]

The failed test has no relationship to this patch. There is probably something else committed to trunk which caused Balancer test failure.

[Hairong Kuang]

The SimulatedDataNodeMetrics seems quite dummy, which is introduced for only the testing purpose.

A side note: DataNodeStatics and DataNodeStaticsMBean collect bytesRead but not bytesWritten. They should collect metrics bytesWritten as well.

[Hairong Kuang]

This patch has the following changes:
1. DataNode's metrics bytesRead & bytesWritten are changed to be long;
2. Add getBytesWritten method to DataNodesStaticsticsMBean;
3. Add a unit test to see if DataNodeMetrics can handle long number of bytes written.

[Sanjay Radia]

+1

[Hadoop QA]

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12391383/metricsLong.patch
against trunk revision 701476.

+1 @author. The patch does not contain any @author tags.

+1 tests included. The patch appears to include 6 new or modified tests.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

+1 Eclipse classpath. The patch retains Eclipse classpath integrity.

-1 core tests. The patch failed core unit tests.

+1 contrib tests. The patch passed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3431/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3431/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3431/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3431/console

This message is automatically generated.

[Hairong Kuang]

This is a patch for branch 18.

[Hairong Kuang]

This is a patch for the trunk. It removes the change to DFSTestUtil that caused the failed tests.

[Brian Bockelman]

Changing the type to Long breaks Ganglia metrics reporting (there's a static mapping of Java type -> Ganglia type; there's no current Long type, meaning the map returns null and causes a NPE).

Ganglia has no Long type - perhaps a solution would be to map Longs in Ganglia to a float?

See also:
HADOOP-4137
HADOOP-3422

[Mac Yang]

> Ganglia has no Long type - perhaps a solution would be to map Longs in Ganglia to a float?

Short of supporting the Long data type, this seems like a reasonable thing to do.

Brian, would you mind coming up with the patch for the Ganglia context?

[Hairong Kuang]

Thank Brian for submitting a patch to HADOOP-4137 for handling Long type in Ganglia context. I am resubmitting my patch to hudson.

[Eric Yang]

+1 for Hairong's patch. The patch provides real test cases.

[Hairong Kuang]

Resubmit the patch since the hudson patch process is back.

[Hadoop QA]

+1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12391556/metricsLong1.patch
against trunk revision 703609.

+1 @author. The patch does not contain any @author tags.

+1 tests included. The patch appears to include 3 new or modified tests.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

+1 Eclipse classpath. The patch retains Eclipse classpath integrity.

+1 core tests. The patch passed core unit tests.

+1 contrib tests. The patch passed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3442/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3442/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3442/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3442/console

This message is automatically generated.

[Hairong Kuang]

I just committed this. Thanks Eric!

[Hudson]

Integrated in Hadoop-trunk #632 (See http://hudson.zones.apache.org/hudson/job/Hadoop-trunk/632/)
. The type of dfs datanoe metrics, bytes_read and bytes_written, is changed to be long. Contributed by Eric Yang and Hairong Kuang.