[
https://issues.apache.org/jira/browse/KAFKA-10649?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Valentina Glagoleva updated KAFKA-10649:
----------------------------------------
Description:
After an upgrade from Kafka 2.2.1 to Kafka 2.5.1 we noticed, that some of our
clusters now use 10-15% more CPU than before:
!CPU_usage_grafana.png|width=613,height=210!
The increase happened right after the rolling upgrade of the cluster and CPU
usage stayed higher than usual since then.
We made no changes to the application side or usage patterns.
The only thing that we noticed is that number of FETCH request seemed to
increase a bit as well. This additional FETCH requests seem to come from
Brokers intercommunication.
What can be the reason for an increase of number of fetch requests and CPU
usage?
was:
After an upgrade from Kafka 2.2.1 to Kafka 2.5.1 we noticed, that some of our
clusters now use 10-15% more CPU than before:
!CPU_usage_grafana.png|width=613,height=210!
The increase happened right after the rolling upgrade of the cluster and CPU
usage stayed higher than usual since then.
We made no changes to the application side or usage patterns.
The only thing that we noticed is that number of FETCH request seem to increase
a bit as well. This additional FETCH requests seem to come from Brokers
intercommunication.
What can be the reason for an increase of number of fetch requests and CPU
usage?
> CPU increase after the upgrade to Kafka 2.5
> -------------------------------------------
>
> Key: KAFKA-10649
> URL: https://issues.apache.org/jira/browse/KAFKA-10649
> Project: Kafka
> Issue Type: Bug
> Components: core
> Affects Versions: 2.5.1
> Environment: Ubuntu 16.04
> Reporter: Valentina Glagoleva
> Priority: Major
> Labels: performance
> Attachments: CPU_usage_grafana.png
>
>
> After an upgrade from Kafka 2.2.1 to Kafka 2.5.1 we noticed, that some of our
> clusters now use 10-15% more CPU than before:
> !CPU_usage_grafana.png|width=613,height=210!
> The increase happened right after the rolling upgrade of the cluster and CPU
> usage stayed higher than usual since then.
> We made no changes to the application side or usage patterns.
> The only thing that we noticed is that number of FETCH request seemed to
> increase a bit as well. This additional FETCH requests seem to come from
> Brokers intercommunication.
> What can be the reason for an increase of number of fetch requests and CPU
> usage?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
