[
https://issues.apache.org/jira/browse/HDFS-16993?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17716655#comment-17716655
]
ASF GitHub Bot commented on HDFS-16993:
---------------------------------------
hfutatzhanghb opened a new pull request, #5597:
URL: https://github.com/apache/hadoop/pull/5597
### Description of PR
In our prod environment, we try to collect datanode metrics every 15s
through jmx_exporter. we found the datanodenetworkerror metric generates a lot.
for example, if we have a cluster with 1000 datanodes, every datanode may
generate 999 datanodenetworkerror metrics, and overall datanodes will generate
1000 multiple 999 = 999000 metrics. This is a very expensive operation. In most
scenarios, we only need the topN of it.
### How was this patch tested?
no need extra UT, TestDataNodeMetrics#testTimeoutMetric has done it.
> Datanode supports configure TopN DatanodeNetworkCounts
> ------------------------------------------------------
>
> Key: HDFS-16993
> URL: https://issues.apache.org/jira/browse/HDFS-16993
> Project: Hadoop HDFS
> Issue Type: Wish
> Affects Versions: 3.3.5
> Reporter: farmmamba
> Priority: Major
>
> In our prod environment, we try to collect datanode metrics every 15s through
> jmx_exporter. we found the datanodenetworkerror metric generates a lot.
> for example, if we have a cluster with 1000 datanodes, every datanode may
> generate 999 datanodenetworkerror metrics, and overall datanodes will
> generate 1000 multiple 999 = 999000 metrics. This is a very expensive
> operation. In most scenarios, we only need the topN of it.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
