/azure-mesos-pokemon

Código asociado al workshop sobre Mesos en Azure.

Primary LanguageHTML

Pokemon Logo

Mesos clusters with Pokémon

Prerequisites

  • You will need the cli in order to interact with Auzre. Install nodejs and then execute npm install -g azure-cli
  • You will also need ssh in your system. If you are a Windows user the easier way to install it is git for Windows.
  • If you don't have a keypair stored in your system you can generate them using ssh-keygen -t rsa -b 2048 -C "email@dominio.com". DON'T OVERWRITE ANY PREVIOUSLY CREATED KEYS.
  • Lastly you can download this project by typing git clone https://github.com/capside/azure-mesos-pokemon.git and cd azure-mesos-pokemon

Cluster creation

  • CLI configuration
azure config mode arm
azure login
  • Ensure your subscription is correctly activated
azure account list
azure account set <subscription_number>
  • If this is your first time with that account you will need to register the needed services
azure provider register --namespace Microsoft.Network
azure provider register --namespace Microsoft.Storage
azure provider register --namespace Microsoft.Compute
azure provider register --namespace Microsoft.ContainerService
  • Check if your quota of CPUs is big enough
azure vm list-usage --location westeurope
  • Define some env variables
set ADMIN_USERNAME=<you_username>
set RESOURCE_GROUP=<a_logic_name>
set DEPLOYMENT_NAME=<name_of_the_deployment>
set ACS_NAME=containerservice-%RESOURCE_GROUP%
set LOCATION=westeurope
set TEMPLATE_URI=https://raw.githubusercontent.com/Azure/azure-quickstart-templates/master/101-acs-dcos/azuredeploy.json
set PARAMFILE=azuredeploy.parameters.json
  • EDIT azuredeploy.parameters.json and set the desired params
  • Deploy the cluster in the resource group
cd azure-arm
azure group create -n %RESOURCE_GROUP% -l %LOCATION% --template-uri %TEMPLATE_URI% -e %PARAMFILE% --deployment-name %DEPLOYMENT_NAME%

azure group deployment show %RESOURCE_GROUP% %DEPLOYMENT_NAME% | grep State

Manage the cluster using the web IU

  • Stablish a ssh tunnel between your laptop and one master
set MASTER=%RESOURCE_GROUP%mgmt.westeurope.cloudapp.azure.com
set AGENTS=%RESOURCE_GROUP%agents.westeurope.cloudapp.azure.com
start ssh -L 80:localhost:80 -N %ADMIN_USERNAME%@%MASTER% -p 2200
  • Open the browser
start http://localhost:80
start http://localhost:80/mesos
start http://localhost:80/marathon

Check the master node (optional)

  • You can check that the master node IP matchs the one shown by the web IU
ssh %ADMIN_USERNAME%@%MASTER% -p 2200
ifconfig | grep "inet addr"

Increase the number of instances in the public VMSS

  • List the deployed VMSS
azure resource list %RESOURCE_GROUP% --resource-type Microsoft.Compute/virtualMachineScaleSets --json
  • Identify the public VMSS (property name, for example "dcos-agent-public-2D554AAB-vmss0")
  • Modify the number of instances on that VMSS
SET PUBLIC_AGENTS_VMSS=<name_of_the_public_VMSS>
azure vmss scale --resource-group %RESOURCE_GROUP% --name %PUBLIC_AGENTS_VMSS% --new-capacity 3

Deploy your pokémon!

  • Take a look to the deployment descriptor in the file deploy-pokemon.json
  • Check the Marathon API (optional)
  • Execute a HTTP order to deploy the application
curl -X POST http://localhost/marathon/v2/apps -d @deploy-pokemon.json -H "Content-type: application/json"

Visualize the deployed containers

  • Check again the Marathon IU
  • Access the application using your browser with http://%AGENTS%:8080
  • Use the API to list the deployed applications
curl -s http://localhost/marathon/v2/apps | prettyjson | grep instances

Scaling containers

  • Scale up the number of containers using the API
curl -X PUT -d "{ \"instances\": 3 }" -H "Content-type: application/json" http://localhost/marathon/v2/apps/poki

Check the resilience of the containers

  • Reload the app in your browser http://%AGENTS%:8080
  • Click over the image of one of the Pokémon
  • Visualize how the container disappears http://localhost/#/nodes/list/
  • In a few seconds a new container should be respawn

Using the DC-OS CLI

  • Download and install the cli from site oficial
  • Configure the API endpoint with dcos config set core.dcos_url http://localhost
  • Try to list the deployed apps with dcos marathon app list

Clean up (IMPORTANT)

azure group delete --name %RESOURCE_GROUP%