[
https://issues.apache.org/jira/browse/KAFKA-13797?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lucas Wang updated KAFKA-13797:
-------------------------------
Description:
It's not a common case, but we experienced the following problem in one of our
clusters.
The use case involves dynamically creating and deleting topics in the cluster,
and the clients were constantly using the special type of Metadata requests
whose topics field is null in order to retrieve all topics before checking a
topic's existence.
A high rate of such Metadata requests generated a heavy load on brokers in the
cluster.
Yet, currently, there is no metric to indicate the metadata response outgoing
bytes rate.
We propose to add such a metric in order to make the troubleshooting of such
cases easier.
was:
It's not a common case, but we experienced the following problem in one of our
clusters.
The use case involves dynamically creating and deleting topics in the cluster,
and the clients were constantly checking if a topic exists in a cluster using
the special type of Metadata requests whose topics field is null in order to
retrieve all topics before checking a topic's existence.
A high rate of such Metadata requests generated a heavy load on brokers in the
cluster.
Yet, currently, there is no metric to indicate the metadata response outgoing
bytes rate.
We propose to add such a metric in order to make the troubleshooting of such
cases easier.
> Adding metric to indicate metadata response outgoing bytes rate
> ---------------------------------------------------------------
>
> Key: KAFKA-13797
> URL: https://issues.apache.org/jira/browse/KAFKA-13797
> Project: Kafka
> Issue Type: Improvement
> Reporter: Lucas Wang
> Priority: Minor
>
> It's not a common case, but we experienced the following problem in one of
> our clusters.
> The use case involves dynamically creating and deleting topics in the
> cluster, and the clients were constantly using the special type of Metadata
> requests whose topics field is null in order to retrieve all topics before
> checking a topic's existence.
> A high rate of such Metadata requests generated a heavy load on brokers in
> the cluster.
> Yet, currently, there is no metric to indicate the metadata response outgoing
> bytes rate.
> We propose to add such a metric in order to make the troubleshooting of such
> cases easier.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
