Question about Sync Interval for Karmada Pull/Push Mode

[CKHuangGH]

Please provide an in-depth description of the question you have:

Hi,

I'd like to know the default interval settings for pulling resource updates and pushing cluster status updates in both the Pull and Push modes of Karmada.

Specifically:

In Pull mode, how frequently does karmada-agent pull new resources from the Karmada API server and push its status back? I have found the setting --cluster-status-update-frequency 10s. It seems this is related to the cluster status rather than pulling new resources.

In Push mode, how often does the Karmada controller manager push the request to the member clusters and fetch status updates from member clusters?

Could you please help me understand the default synchronization intervals and the processes of the Push and Pull in deploying resources? If you can point me to somewhere these settings or discussions can be found?

Thanks!

What do you think about this question?:

Environment:

• Karmada version:
• Kubernetes version:
• Others:

[XiShanYongYe-Chang]

Hi @CKHuangGH, thanks for your question!

In both pull and push clusters, resource propagation and status collection are triggered by events and retries after failures. Therefore, they do not involve time intervals.

The --cluster-status-update-frequency parameter is used to collect cluster status. This is a scheduled task. You can set the interval by using this parameter.

[CKHuangGH]

Hello @XiShanYongYe-Chang,

Thank you for your prompt response. I really appreciate this project because the members are so active.

Your answer is clear to me, and I find this topic very interesting. I would love to learn more about this design.

Could you please clarify the design and process of resource propagation in more detail? Are there any diagrams or documents available? Also, is the agent in the member cluster using an HTTP List-Watch stream?

Additionally, I measured the cross-cluster network traffic between the management cluster and the member cluster. I noticed a sharp increase in network traffic in pull mode and would like to understand the main cause. Could you provide some insights into what kind of data is exchanged in this case? Caused by HTTP List-Watch stream? or what types of cluster status data are get/sent between the management and member cluster?

Thank you for your help!

[XiShanYongYe-Chang]

Could you please clarify the design and process of resource propagation in more detail? Are there any diagrams or documents available?

or https://karmada.io/docs/core-concepts/architecture

Also, is the agent in the member cluster using an HTTP List-Watch stream?

Yes, it is.

I noticed a sharp increase in network traffic in pull mode and would like to understand the main cause. Could you provide some insights into what kind of data is exchanged in this case? Caused by HTTP List-Watch stream? or what types of cluster status data are get/sent between the management and member cluster?

I have not specifically tested this, what is the sudden increase in traffic, is it periodic or only happens once? Did you do anything at that time, such as distributing a large number of resources?