Distributing Worker Load on Different Servers

This documentation provides guidance on using Helm charts to distribute workloads across designated nodes in Kubernetes by leveraging node affinity settings.

In Kubernetes, affinity is a mechanism that allows you to specify rules for scheduling pods on particular nodes based on custom requirements, such as node labels, to optimize resource allocation and performance.

Configuration

Each worker and API component has its own dedicated affinity section in the values.yaml file, allowing you to define custom affinity rules for scheduling. These sections specify node preferences and requirements, enabling you to control the distribution of workloads across different nodes based on specific criteria such as node labels, zone, or instance type.

By configuring affinity settings within each component’s section, you can ensure that workers, API instances, and other services are distributed effectively to meet resource and performance needs.

For example:

• The api.private.nodeAffinity and api.public.nodeAffinity sections control node placement for private and public API instances.
• The worker.background.nodeAffinity, worker.collector.nodeAffinity, and worker.workflow.nodeAffinity sections allow you to set rules for various worker types.

These configurations make it possible to tailor pod deployment based on node characteristics, helping optimize workload management across the Kubernetes cluster.

Affinity-Related Configuration Examples

These are examples for node affinity configuration.

1. API Private Node Affinity (api.private.nodeAffinity):

    # Example of nodeAffinity
    api:
      private:
        nodeAffinity:
          required:
            - key: node-type
              operator: In
              values:
                - compute
          preferred:
            - weight: 100
              key: zone
              operator: In
              values:
                - us-east-1a
            - weight: 50
              key: instance-type
              operator: In
              values:
                - c5.xlarge
   

2. API Public Node Affinity (api.public.nodeAffinity):

    api:
      public:
        nodeAffinity:
          required:
            - key: node-type
              operator: In
              values:
                - compute
          preferred:
            - weight: 100
              key: zone
              operator: In
              values:
                - us-east-1a
            - weight: 50
              key: instance-type
              operator: In
              values:
                - c5.xlarge
   

3. TMS Node Affinity (tms.nodeAffinity):

    tms:
      nodeAffinity:
        ...the same as above
   

4. TMS GUI Node Affinity (tmsGui.nodeAffinity):

    tmsGui:
      nodeAffinity:
        ...the same as above
   

5. Worker Background Node Affinity (worker.background.nodeAffinity): ```yaml worker: background: ...the same as above

6. Worker Collector Node Affinity (worker.collector.nodeAffinity):

    worker:
      collector:
        ...the same as above
   

7. Worker Workflow Node Affinity (worker.workflow.nodeAffinity):

    worker:
      workflow:
        ...the same as above
   

8. APM Profile Node Affinity (apm.profile.nodeAffinity):

    apm:
      profile:
        ...the same as above
   

9. Upgrade Worker Node Affinity (upgrade.nodeAffinity):

    upgrade:
      nodeAffinity:
        ...the same as above
   

10. Queue Bridge Node Affinity (bridge.queue.nodeAffinity):

     bridge:
       queue:
         ...the same as above