Skip to content

Latest commit

 

History

History
257 lines (177 loc) · 9 KB

lab-1-deploying-hybrid-cluster.md

File metadata and controls

257 lines (177 loc) · 9 KB

Multi-Cloud Open DC/OS on AWS with Terraform

AWS and Azure

Requirements

Deploying Multi-Cloud DCOS

This repository is meant to get the bare minimum of running a multi-cloud DC/OS cluster.

This repo is configured to deploy a cluster across AWS and Azure using Cisco CSR 1000V for VPN connection in between.

I. Download Prerequisites

  1. Accept the AWS Cisco CSR subscription from the Marketplace (If you are working from Mesosphere's SE/SA account please skip this step) Click the link below with the same AWS account you will use

https://aws.amazon.com/marketplace/pp?sku=9vr24qkp1sccxhwfjvp9y91p1

  1. Accept the Azure Cisco CSR subscription from the marketplace (If you are working from Mesosphere's SE/SA account please skip this step)
az vm image accept-terms --urn "cisco:cisco-csr-1000v:16_6:16.6.120170804"

  1. Retrieve Sales Mesosphere License Key via OneLogin here: https://mesosphere.onelogin.com/notes/56317. You can also use any valid DC/OS Ent. license key

  2. Retrieve Sales Mesosphere Private and Public Key via OneLogin here: https://mesosphere.onelogin.com/notes/41130. (This is the standard Mesosphere SSH key, You may have this preconfigured already)

  3. Retrieve Mesosphere MAWS Commandline tool for access to AWS: https://github.com/mesosphere/maws/releases

  4. Retrieve Azure CLI tool for access to Azure: https://docs.microsoft.com/en-us/cli/azure/install-azure-cli?view=azure-cli-latest

  5. Retrieve the AWS CLI from here http://docs.aws.amazon.com/cli/latest/userguide/installing.html

  6. Download and install terraform from here https://www.terraform.io/downloads.html . (Or just use brew install terraform)

II. Download the Hybridcloud Demo Terraform Module

mkdir hybridcloud-demo && cd hybridcloud-demo
terraform init -from-module github.com/bernadinm/hybrid-cloud
cp desired_cluster_profile.tfvars.example desired_cluster_profile.tfvars

III. Configure Prerequisites

Configure Mesosphere MAWS

# Download maws-darwin binary from https://github.com/mesosphere/maws/releases
chmod +x maws*
sudo mv ~/Downloads/maws* /usr/local/bin/maws

If you are using maws for the first time, you need to list the account you have access too and login to one of them

Use this command

maws ls

Your output should look something like this:

You will now be taken to your browser for authentication. URL https://aws.mesosphere.com/sso?maws=56014
2 available Accounts:
Team 03:
	110465657741_Mesosphere-PowerUser

Team 10:
	273854932432_Mesosphere-PowerUser

Now use maws to login to retrieve your temporary account credentials

eval $(maws login 110465657741_Mesosphere-PowerUser)

Your output should look something like this

retrieved credentials writing to profile 110465657741_Mesosphere-PowerUser

Note that you will have to refresh the credentials (use the maws login command) every 1 hour

Configure SSH Private and Public Key for Terraform

Set your ssh agent locally to point to your pem key and public key

$ ssh-add /path/to/ssh_private_key.pem

AWS requires you to have an existing keypair set per region. Please set this keypair name so Terraform knows which one to use.

$ cat desired_cluster_profile.tfvars | grep ssh_key_name
ssh_key_name = "<AWS_KEYPAIR_NAME_BY_REGION>"

Azure requires you to provide the public SSH key directly. Please provide this public key so Terraform can pass this to Azure to use.

$ cat desired_cluster_profile.tfvars | grep ssh_pub_key
ssh_pub_key = "<INSERT_SSH_PUB_KEY>"

Note: If any of these entries do not exist in the file, you can add them directly in your desired_cluster_profile.tfvars.

Configure Mesosphere License Key in Terraform variables file

In the desired_cluster_profile.tfvars file in your hybridcloud-demo folder, Copy your license and place it in the dcos_license_key_contents variable

To double check

$ cat desired_cluster_profile.tfvars | grep dcos_license_key_contents
dcos_license_key_contents = "<MY_LICENSE_KEY>"

Configure your aws_profile in Terraform variables file

Copy you Mesosphere maws profile name and provide it to terraform. For the sales team, it is already known to be 110465657741_Mesosphere-PowerUser so it will look like this below:

To double check

$ cat desired_cluster_profile.tfvars | grep aws_profile
aws_profile = "110465657741_Mesosphere-PowerUser"

Configure your Azure login for Terraform

$ az login

Output should be something like

To sign in, use a web browser to open the page https://microsoft.com/devicelogin and enter the code AXUDMJZWB to authenticate.

Note that the code above refreshes every time.

Open the browser link above and enter the code, use your Azure credential to login Once you enter your credential, go back to the terminal. you should see something like

[
  {
    "cloudName": "AzureCloud",
    "id": "6bfddfe6-078b-4a9d-86ff-52e86464efe0",
    "isDefault": true,
    "name": "Mesosphere Sales",
    "state": "Enabled",
    "tenantId": "a98e3fd4-b172-4731-9f98-7038f712e693",
    "user": {
      "name": "[email protected]",
      "type": "user"
    }
  },

IV. Deploy Your First Hybrid Cluster

You are now ready to deploy your hybrid cluster!!

$ terraform apply -var-file desired_cluster_profile.tfvars

Now terraform should start deploying your cluster. This will likely take 10~20 minutes depending on the size of the cluster and the speed of cloud provider region

Here is an output of a successful deployment:

Apply complete! Resources: 114 added, 0 changed, 0 destroyed.

Outputs:

AWS Cisco CSR VPN Router Public IP Address = 18.206.133.140
Azure Cisco CSR VPN Router Public IP Address = 40.118.230.34
Bootstrap Host Public IP = mbernadin-tf80cc-bootstrap.westus.cloudapp.azure.com
Bootstrap Public IP Address = 34.229.179.46
Master ELB Public IP = mbernadin-tf80cc-pub-mas-elb-1841202780.us-east-1.elb.amazonaws.com
Master Public IPs = [
    52.204.155.230
]
Private Agent Public IPs = [
    mbernadin-tf80cc-agent-1.westus.cloudapp.azure.com
]
Public Agent ELB Address = mbernadin-tf80cc-pub-agt-elb-1182904022.us-east-1.elb.amazonaws.com
Public Agent ELB Public IP = public-agent-mbernadin-tf80cc.westus.cloudapp.azure.com
Public Agent Public IPs = [
    mbernadin-tf80cc-public-agent-1.westus.cloudapp.azure.com
]
ssh_user = core

V. Changing AWS Default MTU TCP Value to Match Hybrid Cloud Envs

For hybrid cloud environments, in order for a proper stream of packets to flow between different cloud providers, they must all be ideally using the same MTU. If the source packet is using a higher MTU byte, the packets won't be able to arrive to the desitioation as the packet size is to large to be understood. A good solution here to have this perform in an automated method, which will be coming soon to this repo, is to enable PMTUD (Path MTU Discovery). In the meantime here is a workaround to match all the cloud providers.

On each machine this command can be run:

sudo ip link set dev eth0 mtu 1500

You can verify if it is set by running:

ip a | grep eth0 | grep mtu

If you dont want to log into all the machines, you can run this command via ssh agent forward and it can log into all the machines for you. All the ssh keys needs to be in the ssh-agent for this to work for all host:

Verify Current Setting:

ssh -A core@<master-ip>
for i in $(host slave.mesos | cut -d " " -f 4); do ssh $i -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no  -SsL "ip a | grep eth0 | grep mtu" ;done
for i in $(host master.mesos | cut -d " " -f 4); do ssh $i -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no  -SsL "ip a | grep eth0 | grep mtu" ;done

Set new settings on all machines quickly:

ssh -A core@<master-ip> # skip if you're already on the master
for i in $(host slave.mesos | cut -d " " -f 4); do ssh $i -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no  -SsL "sudo ip link set dev eth0 mtu 1500" ;done
for i in $(host master.mesos | cut -d " " -f 4); do ssh $i -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no  -SsL "sudo ip link set dev eth0 mtu 1500" ;done

You can go back to the command above this one to reverify the changes are made cluster wide.

Destroy Cluster

For the purpose of this lab we will be keeping our cluster up and running, but if you needed to destroy your cluster for any reason now, here is the command:

terraform destroy -var-file desired_cluster_profile.tfvars

Note: No major enhancements should be expected with this repo. It is meant for demo and testing purposes only.

Navigation

  1. LAB1 - Deploying AWS Using Terraform (current)
  2. LAB2 - Bursting from AWS to Azure
  3. LAB3 - Deploying and Migrating Stateless App from AWS to Azure
  4. LAB4 - Deploying Cassandra Multi DataCenter

Return to Main Page