Deployment options on Kubernetes

When you create a site on Kubernetes, there are many options you can use. For example, you can set the number of pods and the resources allocated to each pod. This guide focusses on the following goals:

• Scaling for increased traffic
• Creating a high availability site
• Service synchronization

Scaling for increased traffic

For optimal network latency and throughput, you can adjust the CPU allocation for the router using the router-cpu option. Router CPU is the primary factor governing Skupper network performance.

📌 NOTE
Increasing the number of routers does not improve network performance. An incoming router-to-router link is associated with just one active router. Additional routers do not receive traffic while that router is responding

1. Determine the router CPU allocation you require.

   By default, the router CPU allocation is BestEffort as described in Pod Quality of Service Classes.

   Consider the following CPU allocation options:

2. If you are using the Skupper CLI, set the CPU allocation for the router using the --router-cpu option. For example:

   $ skupper init --router-cpu 2
   

3. If you are using YAML, set the CPU allocation for the router by setting a value for the router-cpu attribute. For example:

   apiVersion: v1
   kind: ConfigMap
   metadata:
     name: "skupper-site"
   data:
     name: "my-site"
     router-cpu: 2
   

Creating a high availability site

By default, Kubernetes restarts any router that becomes unresponsive. (If you encounter router restarts, consider Scaling for increased traffic in order to improve responsiveness.)

If the cluster where you are running Skupper is very busy, it may take time for Kubernetes to schedule a new router pod. You can "preschedule" a backup router by deploying two routers in a site.

1. If you are using the Skupper CLI, set the number of routers to 2 using the --routers option:

   $ skupper init --routers 2
   

2. If you are using YAML, set the number of routers to 2 by setting the routers attribute:

   apiVersion: v1
   kind: ConfigMap
   metadata:
     name: "skupper-site"
   data:
     name: "my-site"
     routers: 2
   

Setting the number of routers to more than two does not provide increased availability and can adversely affect performance.

Note: Clients must reconnect when a router restarts or traffic is redirected to a backup router.

Service synchronization

By default, creating a site enables that site to synchronize all services from other default sites. This means that all services exposed on the service network are available in the current site. For example, if you expose the backend service in the east site, that service is automatically created in the west site.

However, if you want more granular control over which services are available, you can disable service-sync. This might be required if:

• You expose many services and not all are required on all sites.
• You are concerned that a specific service is not available on a specific site.

To disable service synchronization:

$ skupper init --service-sync false

or use the following YAML:

apiVersion: v1
kind: ConfigMap
metadata:
  name: skupper-site
data:
  name: my-site
  service-sync: false

To check whether synchronization is enabled, check the value for service-sync in the output from the following command:

$ kubectl get cm skupper-site -o json

If you disable service-sync and you want to consume an exposed service on a specific site, you can create that service using the following command:

skupper service create <name> <port>

where <name> is the service name on the site where the service is exposed and <port> is the port used to expose that service.

Notes:

• When considering whether services are synchronized between two sites, service-sync must be enabled on both sites.
• If you use the command skupper service delete on a site, that command only works if the service was created on that site.
• Podman sites do not support service-sync.