Getting recommendation on resource requests

[tanken]

hi, i will like to receive cpu & mem recommendation for my containers/pods as it monitors my the workflows, and i thought to use vpa's dry run mode only to realise that the kind: Workflow/CronWorkflow is currently unsupported per this.

was hoping if anyone could recommend any other tools that could provide similar rightsizing recommendations?

[wesleyscholl]

@tanken

You are correct. Currently, the Kubernetes Vertical Pod Autoscaler (VPA) does not support custom Argo Workflows resources like Workflow or CronWorkflow for recommendations. However, there are some alternative approaches:

1. Kubernetes Metrics and Tools:

   • Kubernetes Metrics Server or Prometheus can monitor resource utilization for workflow pods. Based on these metrics, manual resource requests and limits can optimize workflow usage and performance.

   • If Kubernetes Metrics Server is installed:

     • Tools like kubectl top can provide quick insights into real-time CPU and memory usage.

     Example:

     kubectl top pods -n <namespace>

     Example Output:

     NAME                                CPU(cores)   MEMORY(bytes)
     my-pod-1                            1m           20Mi
     my-pod-2                            2m           30Mi
     

2. Third-Party Tools:

   • Tools like Goldilocks (built on VPA), can analyze existing Kubernetes workloads to provide the "right-size" recommendations for CPU and memory requests. Although Goldilocks may not natively support Argo Workflows, it can monitor the underlying Kubernetes pods to determine optimal resource settings.

Resources:

• Kubernetes Tools for Monitoring Resources - https://kubernetes.io/docs/tasks/debug/debug-cluster/resource-usage-monitoring/#resource-metrics-pipeline
• Kubernetes Metrics Server - https://github.com/kubernetes-sigs/metrics-server
• Autoscaling Workloads - https://kubernetes.io/docs/concepts/workloads/autoscaling/
• Vertical Pod Autoscaler - https://github.com/kubernetes/autoscaler/tree/9f87b78df0f1d6e142234bb32e8acbd71295585a/vertical-pod-autoscaler
• Kubernetes Horizontal Pod Autoscaling - https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/
• kubectl - https://kubernetes.io/docs/reference/kubectl/
• Prometheus - https://prometheus.io/
• Goldilocks - https://github.com/FairwindsOps/goldilocks