Best Practices for SkyWalking Architecture for Multi-Cluster Kubernetes

[cahyahikmawan]

Hello, I'm new to using SkyWalking, and I want to make sure I choose the best option for implementing it in my production environment.

For reference, my clusters are:

• infra-tools-cluster
• app-group1-cluster
• app-group2-cluster

I have deployed SkyWalking OAP and SkyWalking UI in the infra-tools-cluster. Everything works perfectly when I deploy a demo app inside infra-tools-cluster. I tested cross-cluster communication with SkyWalking OAP using two methods, both of which worked properly:

1. Assigning a kind: Service to SkyWalking OAP with a LoadBalancer type.
2. Using ingress-nginx with the --tcp-services-configmap flag enabled in the ingress-nginx deployment to proxied port 11800 to Skywalking OAP Service.

Are these good options for running in production? Do you have any recommendations for managing multi-cluster communication? Is SkyWalking Satellite designed for this purpose?

SkyWalking Version: 10.0.1
Elasticsearch Version: 8.14.3
Deployment Mode: Kubernetes (GKE)

Thank you.

[wu-sheng]

What scenarion are you using? Agent, service mesh or something?

[cahyahikmawan]

Thanks for the quick response.

I plan to use the agent to send data to SkyWalking OAP.
Quick question: with a service mesh, do I still need to inject an agent like in Istio?

[wu-sheng]

No, service mesh obs doesn't rely on agent at all.

And how to do load balancer cross clusters is your choice. We don't recommend foe that.

Satellite is also for in cluster balancing.

The important thing is, the agent(Java) is using grpc, which is a keep-connection http2. So, the most load balancer is connection oriented,not data oriented.

[cahyahikmawan]

Okay, I got it. I appreciate your help. 👍 Thanks!