You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Jeff Bean (JIRA)" <ji...@apache.org> on 2011/08/03 19:19:27 UTC
[jira] [Created] (HADOOP-7507) jvm metrics all use the same
namespace
jvm metrics all use the same namespace
--------------------------------------
Key: HADOOP-7507
URL: https://issues.apache.org/jira/browse/HADOOP-7507
Project: Hadoop Common
Issue Type: Bug
Components: metrics
Affects Versions: 0.20.2
Reporter: Jeff Bean
Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099645#comment-13099645 ]
Hudson commented on HADOOP-7507:
--------------------------------
Integrated in Hadoop-Mapreduce-trunk-Commit #856 (See [https://builds.apache.org/job/Hadoop-Mapreduce-trunk-Commit/856/])
HADOOP-7507. Allow ganglia metrics to include the metrics system tags in the gmetric names. Contributed by Alejandro Abdelnur.
todd : http://svn.apache.org/viewcvs.cgi/?root=Apache-SVN&view=rev&rev=1166460
Files :
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/CHANGES.txt
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/conf/hadoop-metrics2.properties
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics/ganglia/GangliaContext.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/sink/ganglia/GangliaSink30.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/metrics2/impl/TestGangliaMetrics.java
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Attachment: HADOOP-7507v1.patch
Patch based on Jeff's changes (thanks Jeff)
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13100273#comment-13100273 ]
Hudson commented on HADOOP-7507:
--------------------------------
Integrated in Hadoop-Hdfs-trunk #787 (See [https://builds.apache.org/job/Hadoop-Hdfs-trunk/787/])
HADOOP-7507. Allow ganglia metrics to include the metrics system tags in the gmetric names. Contributed by Alejandro Abdelnur.
todd : http://svn.apache.org/viewcvs.cgi/?root=Apache-SVN&view=rev&rev=1166460
Files :
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/CHANGES.txt
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/conf/hadoop-metrics2.properties
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics/ganglia/GangliaContext.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/sink/ganglia/GangliaSink30.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/metrics2/impl/TestGangliaMetrics.java
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13091423#comment-13091423 ]
Alejandro Abdelnur commented on HADOOP-7507:
--------------------------------------------
+1.
I've applied 'v2' patch and manually tested with the different versions of ganglia, names are now namespaced. I.e.:
{code}
<METRIC NAME="jvm.NameNode.metrics.logWarn" VAL="0" TYPE="int32" UNITS="" TN="5" TMAX="60" DMAX="0" SLOPE="both">
{code}
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13095617#comment-13095617 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
bq. there is something wrong in the logic, checking.
ProcessName checking and the break is unnecessary. The whole block of logic should look like the code in [this comment|#comment-13092140].
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Todd Lipcon updated HADOOP-7507:
--------------------------------
Resolution: Fixed
Hadoop Flags: [Incompatible change, Reviewed] (was: [Incompatible change])
Status: Resolved (was: Patch Available)
Committed to trunk and 0.23. Should we include this in the 0.20 series as well? Since it's a slightly incompatible change, maybe we need to add a config flag to enable the fixed "jvm" metric in metrics1 in the 0.20 backport?
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13092140#comment-13092140 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
Probably should exclude hostname, as ganglia handles it explicitly:
{code}
for (MetricsTag t : tags) {
if (t.info() != MsInfo.Hostname && t.value() != null) {
sb.append('.').append(t.value());
}
}
{code}
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099640#comment-13099640 ]
Hudson commented on HADOOP-7507:
--------------------------------
Integrated in Hadoop-Common-trunk-Commit #845 (See [https://builds.apache.org/job/Hadoop-Common-trunk-Commit/845/])
HADOOP-7507. Allow ganglia metrics to include the metrics system tags in the gmetric names. Contributed by Alejandro Abdelnur.
todd : http://svn.apache.org/viewcvs.cgi/?root=Apache-SVN&view=rev&rev=1166460
Files :
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/CHANGES.txt
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/conf/hadoop-metrics2.properties
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics/ganglia/GangliaContext.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/sink/ganglia/GangliaSink30.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/metrics2/impl/TestGangliaMetrics.java
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Jeff Bean (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jeff Bean updated HADOOP-7507:
------------------------------
Attachment: JvmMetrics.java
hadoop-metrics.properties
Proposed fix. - a one line change to JvmMetrics.java and a sample hadoop-metrics.properties which includes process-specific metrics.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Attachments: JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Status: Open (was: Patch Available)
there is something wrong in the logic, checking.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Status: Patch Available (was: Open)
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099589#comment-13099589 ]
Todd Lipcon commented on HADOOP-7507:
-------------------------------------
I was thinking the default for the jvm prefix would be uncommented in the file. So, if you enable Ganglia metrics, it's set up correctly by default with no other changes. You could uncomment all of the lines, to be consistent, right? the uncommented line with an empty value would be equivalent to the commented-out line?
For the method being accessible to tests, I think the @VisibleForTests annotation from Guava is more appropriate.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099634#comment-13099634 ]
Todd Lipcon commented on HADOOP-7507:
-------------------------------------
bq. On the annotation, are we using Guava annotations for that in Hadoop?
Not a lot, yet - I've used it a few places in HDFS, but that's about it. I'm fine with leaving it as you have it.
bq. Can we defer this to a follow up tune-up JIRA?
Oh, I see what you're saying. Let's defer to a followup once metrics2 starts getting real usage, like you said.
+1, I'll commit this momentarily
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Todd Lipcon updated HADOOP-7507:
--------------------------------
Release Note: JVM metrics published to Ganglia now include the process name as part of the gmetric name.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Status: Open (was: Patch Available)
patch has to be rebased to new layout. And there is concatenation typo.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13092143#comment-13092143 ]
Todd Lipcon commented on HADOOP-7507:
-------------------------------------
We'd also want a consistent ordering of tags, right? (perhaps sorted by their keys)
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13093062#comment-13093062 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
v4 patch integrated generic tags handling in jvm context only :) i.e., the outer if check for jvm context is unnecessary.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Fix Version/s: 0.23.0
Status: Patch Available (was: Open)
rebased and concat typo fixed
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13095506#comment-13095506 ]
Hadoop QA commented on HADOOP-7507:
-----------------------------------
-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12492641/HADOOP-7507v5.patch
against trunk revision .
+1 @author. The patch does not contain any @author tags.
-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no new tests are needed for this patch.
Also please list what manual steps were performed to verify this patch.
+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 (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
-1 core tests. The patch failed these unit tests:
org.apache.hadoop.io.serializer.avro.TestAvroSerialization
org.apache.hadoop.io.TestText
org.apache.hadoop.conf.TestConfiguration
org.apache.hadoop.io.TestEnumSetWritable
org.apache.hadoop.fs.TestPath
org.apache.hadoop.ipc.TestAvroRpc
+1 contrib tests. The patch passed contrib unit tests.
Test results: https://builds.apache.org/job/PreCommit-HADOOP-Build/121//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/121//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-annotations.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/121//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth-examples.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/121//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-common.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/121//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth.html
Console output: https://builds.apache.org/job/PreCommit-HADOOP-Build/121//console
This message is automatically generated.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Jeff Bean (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jeff Bean updated HADOOP-7507:
------------------------------
Attachment: screenshot-1.jpg
This screen shot shows how process-specific jvm metrics can appear in ganglia.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Attachments: JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13095613#comment-13095613 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
bq. generic tag handling won't work well in Ganglia, given that the JT has job-specific metrics tagged by the job ID (eg fairscheduler job level metrics)
Ganglia cannot handle this type of metrics correctly anyway. metrics2 can be configured (dynamically) to filter out certain portion of metrics (by context/source/metrics/tag with glob pattern and regex) per metrics sink.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Attachment: HADOOP-7507v6.patch
v6,
For 'metrics' the original fix stays, just prefixing the processName to the 'jvm' context.
For 'metrics2' rewritten (as per Todd's suggestion) to be able to specify which tags are to be used for the prefix.
In the hadoop-metrics2.properties the property *PROCESS_NAME.sink.ganglia.tagsForPrefix.CONTEXT_NAME* indicates which tags are used. If not specified, no tag is used, if '*' is specified all tags are used, individual tags can be specified separated by commas.
For example:
{code}
namenode.sink.ganglia.tagsForPrefix.jvm=*
namenode.sink.ganglia.tagsForPrefix.dfs=*
{code}
The Context and Hostname tags are never used.
Because the default is no tags, the default behavior is backwards compatible.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099596#comment-13099596 ]
Hadoop QA commented on HADOOP-7507:
-----------------------------------
+1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12493507/HADOOP-7507v7.patch
against trunk revision .
+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 (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
+1 core tests. The patch passed unit tests in hadoop-common-project.
+1 contrib tests. The patch passed contrib unit tests.
Test results: https://builds.apache.org/job/PreCommit-HADOOP-Build/148//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/148//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/148//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-common.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/148//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-annotations.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/148//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth-examples.html
Console output: https://builds.apache.org/job/PreCommit-HADOOP-Build/148//console
This message is automatically generated.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099641#comment-13099641 ]
Hudson commented on HADOOP-7507:
--------------------------------
Integrated in Hadoop-Hdfs-trunk-Commit #922 (See [https://builds.apache.org/job/Hadoop-Hdfs-trunk-Commit/922/])
HADOOP-7507. Allow ganglia metrics to include the metrics system tags in the gmetric names. Contributed by Alejandro Abdelnur.
todd : http://svn.apache.org/viewcvs.cgi/?root=Apache-SVN&view=rev&rev=1166460
Files :
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/CHANGES.txt
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/conf/hadoop-metrics2.properties
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics/ganglia/GangliaContext.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/sink/ganglia/GangliaSink30.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/metrics2/impl/TestGangliaMetrics.java
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13095524#comment-13095524 ]
Todd Lipcon commented on HADOOP-7507:
-------------------------------------
Thinking about this a bit... it seems like doing the generic tag handling won't work well in Ganglia, given that the JT has job-specific metrics tagged by the job ID (eg fairscheduler job level metrics). Over time, the JT will accumulate an unbounded number of metrics in ganglia, and I don't think Ganglia ever "ages off" a gmetric that it hasn't seen in a while. I think we need some sort of config for the ganglia context that says which tag names need to be included as part of a metric (and perhaps which metrics shouldn't be reported at all in ganglia?)
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13084486#comment-13084486 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
This is actually not the right fix as jvm metrics already set the processName tag. The right place to fix this is actually in ganglia plugin to encode tags as part of the metrics names, as the problem not only happens to jvm metrics but also other per tag metrics as well, say per queue metrics in capacity scheduler.
Both metrics1 and metrics2 Ganglia{Plugin,Sink} needs to be fixed. Metrics2 is what hadoop will be using for trunk/0.23. Some downstream projects like HBase will still use metrics1 for compatibility reasons (working with multiple hadoop releases from 0.20 to 0.23 etc.)
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Attachment: HADOOP-7507v3.patch
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13085551#comment-13085551 ]
Hadoop QA commented on HADOOP-7507:
-----------------------------------
-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12490506/HADOOP-7507-v2.patch
against trunk revision .
+1 @author. The patch does not contain any @author tags.
-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no new tests are needed for this patch.
Also please list what manual steps were performed to verify this patch.
+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 (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
+1 core tests. The patch passed core unit tests.
+1 contrib tests. The patch passed contrib unit tests.
+1 system test framework. The patch passed system test framework compile.
Test results: https://builds.apache.org/job/PreCommit-HADOOP-Build/40//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/40//artifact/trunk/target/newPatchFindbugsWarningshadoop-common.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/40//artifact/trunk/target/newPatchFindbugsWarningshadoop-annotations.html
Console output: https://builds.apache.org/job/PreCommit-HADOOP-Build/40//console
This message is automatically generated.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13100292#comment-13100292 ]
Hudson commented on HADOOP-7507:
--------------------------------
Integrated in Hadoop-Mapreduce-trunk #810 (See [https://builds.apache.org/job/Hadoop-Mapreduce-trunk/810/])
HADOOP-7507. Allow ganglia metrics to include the metrics system tags in the gmetric names. Contributed by Alejandro Abdelnur.
todd : http://svn.apache.org/viewcvs.cgi/?root=Apache-SVN&view=rev&rev=1166460
Files :
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/CHANGES.txt
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/conf/hadoop-metrics2.properties
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics/ganglia/GangliaContext.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/metrics2/sink/ganglia/GangliaSink30.java
* /hadoop/common/trunk/hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/metrics2/impl/TestGangliaMetrics.java
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13096358#comment-13096358 ]
Hadoop QA commented on HADOOP-7507:
-----------------------------------
-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12492679/HADOOP-7507v6.patch
against trunk revision .
+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 (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
-1 core tests. The patch failed these unit tests:
org.apache.hadoop.io.serializer.avro.TestAvroSerialization
org.apache.hadoop.io.TestText
org.apache.hadoop.conf.TestConfiguration
org.apache.hadoop.io.TestEnumSetWritable
org.apache.hadoop.fs.TestPath
org.apache.hadoop.ipc.TestAvroRpc
+1 contrib tests. The patch passed contrib unit tests.
Test results: https://builds.apache.org/job/PreCommit-HADOOP-Build/127//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/127//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-annotations.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/127//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth-examples.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/127//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-common.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/127//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth.html
Console output: https://builds.apache.org/job/PreCommit-HADOOP-Build/127//console
This message is automatically generated.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Michael Katzenellenbogen (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Michael Katzenellenbogen updated HADOOP-7507:
---------------------------------------------
Attachment: HADOOP-7507-v2.patch
Ran into this issue as well. Haven't fully tested the patch yet (functionally), but uploading for reference.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13092115#comment-13092115 ]
Todd Lipcon commented on HADOOP-7507:
-------------------------------------
+1 pending Hudson
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Attachment: HADOOP-7507v5.patch
v5 patch addressing last comment from Luke.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Todd Lipcon (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099229#comment-13099229 ]
Todd Lipcon commented on HADOOP-7507:
-------------------------------------
- In hadoop-metrics2.properties, we should probably default tagsForPrefix.jvm=ProcessName, right? So that the default behavior with the new metrics system solves the original issue of this JIRA
- TAGS_FOR_PREFIX_PROPERTY can be private
- minor efficiency nit: change createPrefix() to be appendPrefix(MetricsRecord record, StringBuilder builder) so you can just call appendPrefix(record, sb);
- createPrefix could be package-private, right?
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13092983#comment-13092983 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
The ordering of the tags is determined by the metrics sources (typically managed by MetricsRegistry, which uses a LinkedHashMap to manage tags). The same metrics record with different tags are from different instances of the same metrics classes. So sorting is not necessary. I personally would prefer being able control the ordering.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Luke Lu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13092139#comment-13092139 ]
Luke Lu commented on HADOOP-7507:
---------------------------------
The patch only works around the jvm metrics problem. It doesn't really fix the underlying issue of the Ganglia support: lack of tag/dimension/group support (They are [working on tags support|http://www.mail-archive.com/ganglia-general@lists.sourceforge.net/msg06822.html]). Our current workaround is flatten the namespace, but in a inconsistent way, so the it still won't work for QueueMetrics in CapacityScheduler.
I think for metrics1, the hardcoded workaround is probably OK (to minimize disruption to old systems) but for metrics2, we could do something like this for the groupName:
{code}
for (MetricsTag t : tags) {
if (t.value() != null) {
sb.append('.').append(t.value());
}
}
{code}
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13099611#comment-13099611 ]
Alejandro Abdelnur commented on HADOOP-7507:
--------------------------------------------
On the annotation, are we using Guava annotations for that in Hadoop?
Regarding uncommenting the *.sink.ganglia.tagsForPrefix.jvm=, that is a good idea, but what are the '*' in the new world (MR2)? Can we defer this to a follow up tune-up JIRA?
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Attachment: HADOOP-7507v7.patch
v7 addresses Todd's comments with a minor twist, using @InterfaceAudience.Private annotation because the testcase is in a different package.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, HADOOP-7507v7.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13093067#comment-13093067 ]
Hadoop QA commented on HADOOP-7507:
-----------------------------------
-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12492109/HADOOP-7507v4.patch
against trunk revision .
+1 @author. The patch does not contain any @author tags.
-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no new tests are needed for this patch.
Also please list what manual steps were performed to verify this patch.
+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 appears to introduce 4 new Findbugs (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
-1 core tests. The patch failed these unit tests:
+1 contrib tests. The patch passed contrib unit tests.
Test results: https://builds.apache.org/job/PreCommit-HADOOP-Build/96//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/96//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-annotations.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/96//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-common.html
Findbugs warnings: https://builds.apache.org/job/PreCommit-HADOOP-Build/96//artifact/trunk/patchprocess/newPatchFindbugsWarningshadoop-auth.html
Console output: https://builds.apache.org/job/PreCommit-HADOOP-Build/96//console
This message is automatically generated.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Fix Version/s: 0.24.0
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0, 0.24.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, HADOOP-7507v5.patch, HADOOP-7507v6.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Status: Patch Available (was: Open)
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: HADOOP-7507v1.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur updated HADOOP-7507:
---------------------------------------
Attachment: HADOOP-7507v4.patch
v4 patch.
Integrates Luke's suggestions.
I've checked with Matt regarding the exclusion of the Hostname tag, and he indicated that it is a good idea. His argument is that hostnames are case insensitive and different DNSes may have the host with different case and this could create shadow groups if a process starts getting a different case DNS name.
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Fix For: 0.23.0
>
> Attachments: HADOOP-7507-v2.patch, HADOOP-7507v1.patch, HADOOP-7507v3.patch, HADOOP-7507v4.patch, JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Assigned] (HADOOP-7507) jvm metrics all use the same
namespace
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-7507?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur reassigned HADOOP-7507:
------------------------------------------
Assignee: Alejandro Abdelnur
> jvm metrics all use the same namespace
> --------------------------------------
>
> Key: HADOOP-7507
> URL: https://issues.apache.org/jira/browse/HADOOP-7507
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 0.20.2
> Reporter: Jeff Bean
> Assignee: Alejandro Abdelnur
> Attachments: JvmMetrics.java, hadoop-metrics.properties, screenshot-1.jpg
>
>
> Ganglia jvm metrics don't make sense because it's not clear which java process the metrics refer to. In fact, all hadoop java processes running on a node report their jvm metrics to the same namespace.
> The metrics are exposed by the "jvm" context in JvmMetrics.java. This leads to confusing and nonsensical graphs in ganglia and maybe other monitoring tools.
> One way to fix this is to make sure the process name is reported in the jvm context, making it clear which process is associated with the context, and separating out the jvm metrics per process.
> This is marked as an "incompatible change" because the fix provided removes the JVM metrics and replaces it with process-specific metrics.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira