[
https://issues.apache.org/jira/browse/HBASE-24994?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Javier Akira Luca de Tena updated HBASE-24994:
----------------------------------------------
Description:
When adding Hedged Reads metrics at
https://issues.apache.org/jira/browse/HBASE-12220, hedgedReadOpsInCurThread was
not included.
I wonder if there was any reason to not include them. If not, I will provide
pull request to include them soon.
Hedged reads use same thread pool for the original read:
[https://github.com/apache/hadoop/blob/0b8464d75227fcee2c6e7f2410377b3d53d3d5f8/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSInputStream.java#L1349-L1357]
When the executor is full and rejects the task, it will be executed in current
thread, increasing this metric:
[https://github.com/apache/hadoop/blob/0b8464d75227fcee2c6e7f2410377b3d53d3d5f8/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSClient.java#L2913-L2918]
This metric is very useful to determine if the hedged reads thread pool
(dfs.client.hedged.read.threadpool.size) is undersized.
was:
When adding Hedged Reads metrics at
https://issues.apache.org/jira/browse/HBASE-12220, hedgedReadOpsInCurThread was
not included.
I wonder if there was any reason to not include them. If not, I will provide
pull request to include them soon.
Hedged reads use same thread pool for the original read:
[https://github.com/apache/hadoop/blob/0b8464d75227fcee2c6e7f2410377b3d53d3d5f8/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSInputStream.java#L1349-L1357]
When the executor is full and rejects the task, it will be executed in current
thread, increasing this metric:
[https://github.com/apache/hadoop/blob/0b8464d75227fcee2c6e7f2410377b3d53d3d5f8/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSClient.java#L2913-L2918]
This metric is very useful the determine if the hedged reads thread pool
(dfs.client.hedged.read.threadpool.size) is undersized.
> Add hedgedReadOpsInCurThread metric
> -----------------------------------
>
> Key: HBASE-24994
> URL: https://issues.apache.org/jira/browse/HBASE-24994
> Project: HBase
> Issue Type: Improvement
> Components: metrics
> Reporter: Javier Akira Luca de Tena
> Assignee: Javier Akira Luca de Tena
> Priority: Minor
>
> When adding Hedged Reads metrics at
> https://issues.apache.org/jira/browse/HBASE-12220, hedgedReadOpsInCurThread
> was not included.
> I wonder if there was any reason to not include them. If not, I will provide
> pull request to include them soon.
>
> Hedged reads use same thread pool for the original read:
>
> [https://github.com/apache/hadoop/blob/0b8464d75227fcee2c6e7f2410377b3d53d3d5f8/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSInputStream.java#L1349-L1357]
> When the executor is full and rejects the task, it will be executed in
> current thread, increasing this metric:
>
> [https://github.com/apache/hadoop/blob/0b8464d75227fcee2c6e7f2410377b3d53d3d5f8/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSClient.java#L2913-L2918]
> This metric is very useful to determine if the hedged reads thread pool
> (dfs.client.hedged.read.threadpool.size) is undersized.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
