[
https://issues.apache.org/jira/browse/ZOOKEEPER-5015?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Li Wang updated ZOOKEEPER-5015:
-------------------------------
Description:
Implement a new admin server command (i.e. {{shed_connections)}} that allows
operators to drop a specified percentage of client connections for load
management and operational control.
Key Use Cases:
* *Fixing client connection imbalance* caused by rolling restarts or server
failures, where connections concentrate on remaining servers
* *Preventing thundering herd effects* during planned server maintenance by
gradually redistributing connections before taking a server offline
* *Enabling proportional load reduction* on overloaded servers (e.g., leaders
experiencing high CPU/memory usage) to maintain cluster stability
was:
Implement a new admin server command (i.e. {{shed_connections)}} that allows
operators to drop a specified percentage of client connections for load
management and operational control.
Key Use Cases:
* *Fixing client connection imbalance* caused by rolling restarts or server
failures, where connections concentrate on remaining servers
* *Preventing thundering herd effects* during planned server maintenance by
gradually redistributing connections before taking a server offline
* *Enabling proportional load reduction* on overloaded servers (e.g., leaders
experiencing high CPU/memory usage) to maintain cluster stability
> Add admin server command to shed client connections by percentage
> -----------------------------------------------------------------
>
> Key: ZOOKEEPER-5015
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-5015
> Project: ZooKeeper
> Issue Type: New Feature
> Components: server
> Affects Versions: 3.10.0
> Reporter: Li Wang
> Priority: Major
>
> Implement a new admin server command (i.e. {{shed_connections)}} that allows
> operators to drop a specified percentage of client connections for load
> management and operational control.
> Key Use Cases:
> * *Fixing client connection imbalance* caused by rolling restarts or server
> failures, where connections concentrate on remaining servers
> * *Preventing thundering herd effects* during planned server maintenance by
> gradually redistributing connections before taking a server offline
> * *Enabling proportional load reduction* on overloaded servers (e.g.,
> leaders experiencing high CPU/memory usage) to maintain cluster stability
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
