Two scenarios of multi-cluster management, and references to the corresponding documentation - Alibaba Cloud Service Mesh

Prerequisite concepts

Disaster recovery unit

In disaster recovery scenarios, to avoid single points of failure, a complete, independently operating system is deployed in each of multiple independent geographic locations. Each such single system is called a disaster recovery unit.

Flat network

Consider two independent subnets (for example, different Alibaba Cloud virtual private clouds (VPCs) or vSwitches). If the IP addresses of one subnet can be directly accessed from the other subnet, the two subnets form a flat network. Direct access means that Layer 3 (IP layer) routing can reach the other subnet directly. Otherwise, they form a non-flat network.

Typical multi-cluster scenarios

The multi-cluster capabilities of Service Mesh (ASM) support various multi-cluster deployment solutions.

Scenario 1: Multi-cluster disaster recovery

When you build multiple clusters for disaster recovery purposes, the clusters should be deployed in multiple geographic locations. In addition, you should create a dedicated ASM instance for each cluster in the same geographic location as the cluster. In multi-cloud scenarios, if you use clusters from other cloud providers and Alibaba Cloud is not available in the corresponding region, select the nearest geographic location. For more information, see Multi-cluster disaster recovery.

Scenario 2: Multi-cluster shared service discovery

Suppose your workloads are asymmetrically deployed (that is, different workloads are deployed in each cluster) across multiple clusters within a single disaster recovery unit (for example, the same zone or the same region). If applications across the clusters may call each other, you can add these clusters to the same ASM instance. The clusters then share service discovery information, and applications can directly call each other across clusters by using Kubernetes service domain names, without the need to expose services outside the clusters through gateways.

Multi-cluster network connectivity

In the preceding multi-cluster scenarios, cross-cluster access may be required. For example, in disaster recovery scenarios, when the local cluster becomes unavailable, cross-cluster calls are attempted. The clusters may reside in different networks, for example, across VPCs, across regions, or in hybrid cloud environments where networks are not interconnected. In this case, you must first establish network connections between the clusters. The following methods are available:

Deployment type	Description	Connection method
Cross VPCs on Alibaba Cloud	All clusters are Alibaba Cloud clusters, but they reside in different VPCs	ASM cross-cluster mesh proxy over the Internet: This method does not depend on other cloud products, but is limited by Internet quality. It is suitable for all scenarios. For more information, see Use ASM cross-cluster mesh proxy to implement cross-network communication among multiple clusters Cloud Enterprise Network (CEN): CEN is a paid product, but provides better network quality. It applies to scenarios where cluster VPCs reside in different regions. For more information, see Secure traffic exchange with an Enterprise transit router to connect the VPC networks.
Alibaba Cloud + third-party cloud or self-managed clusters	The deployment includes Alibaba Cloud clusters and third-party cloud or self-managed clusters	ASM cross-cluster mesh proxy over the Internet: This method does not depend on other cloud products, but is limited by Internet quality. It is suitable for all scenarios. For more information, see Use ASM cross-cluster mesh proxy to implement cross-network communication among multiple clusters. Leased lines: In general, the cost is higher, but the network quality is better.

Multi-cluster traffic management

Use a DNS proxy for multi-cluster service discovery

Multi-cluster disaster recovery

Region-level disaster recovery with ASM and GTM

Using a service mesh for availability zone disaster recovery

Use a service mesh for service-level disaster recovery

Multi-cluster observability

Enable Mesh Topology in Managed Mode

Others

Manage multi-cluster applications with ASM and Karmada