Skip to content

Latest commit

 

History

History
125 lines (94 loc) · 4.04 KB

eks.md

File metadata and controls

125 lines (94 loc) · 4.04 KB

Prombench in EKS

Run prombench tests in Elastic Kubernetes Service.

Setup prombench

  1. Create the main node
  2. Deploy monitoring components

Create the Main Node

accesskeyid: <Amazon access key>
secretaccesskey: <Amazon access secret>
  • Create a VPC with public subnets.
  • Create a Amazon EKS cluster role with following policies:
    • AmazonEKSclusterPolicy
  • Create a Amazon EKS worker node role with following policies:
    • AmazonEKSWorkerNodePolicy, AmazonEKS_CNI_Policy, AmazonEC2ContainerRegistryReadOnly
  • Set the following environment variables and deploy the cluster.
export AUTH_FILE=<path to yaml credentials file that was created in the last step>
export CLUSTER_NAME=prombench
export ZONE=us-east-1
export EKS_WORKER_ROLE_ARN=<Amazon EKS worker node IAM role ARN>
export EKS_CLUSTER_ROLE_ARN=<Amazon EKS cluster role ARN>
# By default SEPARATOR DeploymentVar is set to `,` but you can override it by exporting and
# then passing it with the -v flag. It is used to split DeploymentVar into a slice.
export SEPARATOR=, 
export EKS_SUBNET_IDS=SUBNETID1,SUBNETID2,SUBNETID3
export PROVIDER=eks

make cluster_create

Deploy monitoring components

Collecting, monitoring and displaying the test results and logs

  • [Optional] If used with the Github integration generate a GitHub auth token. 
export GRAFANA_ADMIN_PASSWORD=password
export DOMAIN_NAME=prombench.prometheus.io // Can be set to any other custom domain or an empty string when not used with the Github integration.
export OAUTH_TOKEN=<generated token from github or set to an empty string " ">
export WH_SECRET=<github webhook secret>
export GITHUB_ORG=prometheus
export GITHUB_REPO=prometheus
make cluster_resource_apply
  • The output will show the ingress IP which will be used to point the domain name to.
  • Set the A record for <DOMAIN_NAME> to point to nginx-ingress-controller IP address.
  • The services will be accessible at:
    • Grafana :: http://<DOMAIN_NAME>/grafana
    • Prometheus :: http://<DOMAIN_NAME>/prometheus-meta
    • Logs :: http://<DOMAIN_NAME>/grafana/explore
    • Profiles :: http://<DOMAIN_NAME>/profiles

Usage

Start a benchmarking test manually

  • Set the following environment variables.
export RELEASE=<master/main or any prometheus release(ex: v2.3.0) >
export PR_NUMBER=<PR to benchmark against the selected $RELEASE>
  • Create the nodegroups for the k8s objects
make node_create
  • Deploy the k8s objects
make resource_apply

Stopping a benchmarking test manually

  • Set the following environment variables:
export AUTH_FILE=<path to yaml credentials file that was created>
export CLUSTER_NAME=prombench
export SEPARATOR=,
export EKS_WORKER_ROLE_ARN=<Amazon EKS worker node IAM role ARN>
export EKS_CLUSTER_ROLE_ARN=<Amazon EKS cluster role ARN>
export EKS_SUBNET_IDS=SUBNETID1,SUBNETID2,SUBNETID3
export ZONE=us-east-1
export PROVIDER=eks

export PR_NUMBER=<PR to benchmark against the selected $RELEASE>
  • To delete just the nodepool (while keeping the cluster's main node intact), run:
make clean
  • To delete everything (complete teardown of the entire cluster and all the resources), run:
make cluster_delete