[
https://issues.apache.org/jira/browse/HBASE-29616?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eungsop Yoo updated HBASE-29616:
--------------------------------
Description:
RegionServers in HBase can sometimes experience abnormally high load, such as
excessive CPU usage or network traffic. Existing methods to diagnose the cause
include (1) using metrics, which are only available at the Region level, (2)
checking recent ResponseTooSlow / ResponseTooLarge logs, which provide only
limited visibility, and (3) employing third-party tools like
[hbase-packet-inspector|https://github.com/kakao/hbase-packet-inspector], which
enable detailed analysis but cannot guarantee complete request tracking and
require external management.
This PR proposes adding a lightweight RequestStatsCollector inside the
RegionServer to capture all requests at the row level in real time. The
collected statistics can be examined and analyzed through the RegionServer web
console.
For more details, please refer to the following document.
https://docs.google.com/document/d/1ELfSwT8d1ctE7QWJvRX1YTVpcVnIvGFhM0MczuiXg6E/edit?tab=t.0
https://github.com/apache/hbase/pull/7334
was:
RegionServers in HBase can sometimes experience abnormally high load, such as
excessive CPU usage or network traffic. Existing methods to diagnose the cause
include (1) using metrics, which are only available at the Region level, (2)
checking recent ResponseTooSlow / ResponseTooLarge logs, which provide only
limited visibility, and (3) employing third-party tools like
[hbase-packet-inspector|https://github.com/kakao/hbase-packet-inspector], which
enable detailed analysis but cannot guarantee complete request tracking and
require external management.
This PR proposes adding a lightweight RequestStatsCollector inside the
RegionServer to capture all requests at the row level in real time. The
collected statistics can be examined and analyzed through the RegionServer web
console.
For more details, please refer to the following document.
https://docs.google.com/document/d/1ELfSwT8d1ctE7QWJvRX1YTVpcVnIvGFhM0MczuiXg6E/edit?tab=t.0
> Add Real-time Request Stats Collector
> -------------------------------------
>
> Key: HBASE-29616
> URL: https://issues.apache.org/jira/browse/HBASE-29616
> Project: HBase
> Issue Type: New Feature
> Components: monitoring
> Reporter: Eungsop Yoo
> Assignee: Eungsop Yoo
> Priority: Major
> Labels: pull-request-available
>
> RegionServers in HBase can sometimes experience abnormally high load, such as
> excessive CPU usage or network traffic. Existing methods to diagnose the
> cause include (1) using metrics, which are only available at the Region
> level, (2) checking recent ResponseTooSlow / ResponseTooLarge logs, which
> provide only limited visibility, and (3) employing third-party tools like
> [hbase-packet-inspector|https://github.com/kakao/hbase-packet-inspector],
> which enable detailed analysis but cannot guarantee complete request tracking
> and require external management.
> This PR proposes adding a lightweight RequestStatsCollector inside the
> RegionServer to capture all requests at the row level in real time. The
> collected statistics can be examined and analyzed through the RegionServer
> web console.
> For more details, please refer to the following document.
> https://docs.google.com/document/d/1ELfSwT8d1ctE7QWJvRX1YTVpcVnIvGFhM0MczuiXg6E/edit?tab=t.0
> https://github.com/apache/hbase/pull/7334
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
