Skip to content

Latest commit

 

History

History
88 lines (59 loc) · 3.05 KB

HOW_TO_RUN.md

File metadata and controls

88 lines (59 loc) · 3.05 KB
Raw

Pre-requisites:

  • Minimum 3-node Kubernetes cluster
  • Portworx installed on the Kubernetes cluster
  • Torpedo currently requires root username/password to access the physical nodes.

Running using a Kubernetes pod

Kubernetes with ssh node driver

To run torpedo with ssh node driver run the following script:

$ TORPEDO_SSH_PASSWORD=[your_root_password] deployments/deploy-ssh.sh

Kubernetes with aws node driver

To run torpedo under AWS environment with aws node driver run the following script:

$ deployments/deploy-aws.sh

Make sure you change image: portworx/torpedo:latest to your torpedo docker image.

The above command starts Torpedo by deploying a k8s Pod in your kubernetes cluster. It also specified Portworx (pxd) as the volume driver and ssh as the node driver to.

You can look at status of torpedo by viewing logs using: kubectl logs -f torpedo

Running directly using ginkgo

First expose KUBECONFIG to torpedo can talk to the k8s API.

export KUBECONFIG=<location_of_k8s_cluster_kubeconfig_file>

It is highly suggested to run your new test against a single app

For example: ginkgo --focus SetupTeardown -v bin/basic.test -- -spec-dir `pwd`/drivers/scheduler/k8s/specs --app-list elasticsearch

This will run the basic test against just elastic search

To run a subset of tests use ginkgo's --focus parameter: ginkgo --focus "AppTasksDown|VolumeDriverDown" -v bin/basic.test -- -spec-dir `pwd`/drivers/scheduler/k8s/specs

To run all tests: ginkgo -v bin/basic.test -- -spec-dir `pwd`/drivers/scheduler/k8s/specs

To dry-run all tests: ginkgo -dryRun -v bin/*.test -- -spec-dir `pwd`/drivers/scheduler/k8s/specs

Running torpedo on EKS

NOTE: perform the steps below if cluster was created using eksctl or any tool, other than dedicated eks jenkins job

(if cluster was not provisioned with spawn)

  1. Provision a cluster of nodes n+1 (more than you need)

  2. Create a "master" node. Torpedo contains a test which reboots all px nodes, hence it needs to be alive. We don't have a master node on EKS, so we just simulate it :

    1. Pick a node
    2. disable PX on it
    3. taint
    kubectl get nodes '--output=jsonpath={.items[0].metadata.name}'

kubectl label node <node_name> px/enabled=false

kubectl taint node <node_name> apps=false:NoSchedule

  3. Install PX in a usual way

  4. Add the following parameters to the [deploy-ssh.sh](deployments/deploy-ssh.sh) script at the very beginning of the file

#KUBECONFIG=/tmp/kubeconfig # if you don't specify the config, torpedo will pick up the system config
TORPEDO_IMG=portworx/torpedo:master # override with your image if needed
VERBOSE=true
FOCUS_TESTS=SetupTeardown,AppTasksDown,VolumeDriverDown,VolumeDriverAppDown,VolumeDriverDownAttachedNode,VolumeDriverCrash,AppScaleUpAndDown,VolumeUpdate
SCALE_FACTOR=1
FAIL_FAST=true
APP_LIST=postgres,sysbench,nginx-sharedv4,vdbench-sharedv4
K8S_VENDOR=eks

  1. Add your test to the FOCUS_TESTS: FOCUS_TESTS=SetupTeardown,AppTasksDown,...,<YOUR_TEST_NAME>

  2. Run ./deploy-ssh.sh