Skip to content

Latest commit

 

History

History
162 lines (130 loc) · 4.13 KB

README.md

File metadata and controls

162 lines (130 loc) · 4.13 KB
Raw

infrastructure/sysops

Peg.ski runs on AWS.

The AWS infrastructure is maintained by Terraform. Services are being deployed using ansible.

required

To be able to setup and provision the peg.ski infrastructure you need to have installed:

  • terraform (> 0.7)
  • ansible (>2.1)

For creating maps of the terraformed infrastructure you will need:

  • Graphviz (>)

besides that you will need AWS AIM account credentials and the deploy user private key.

secrets

You will need to provide AWS credentials as environment variables in order to be able to run the scripts.

These variables can be added to env/environment.local (for all linux versions) or environment.osx for OS X. The following command can be used to load the environment in your current working shell:

cd env; source ./loadenv.sh; cd -

Additionally, you will need the ssh public and private key of the deploy user for terraform. These should be placed in terraform/ssh.

terraform state

The statefile for terraform is saved in a dedicated S3 bucket. For this S3 bucket we applied the following S3 policy:

{
   "Version": "2012-10-17",
   "Statement": [
      {
         "Sid": "statement1",
         "Effect": "Allow",
         "Principal": {
            "AWS": "arn:aws:iam::<insert-arn-number>:user/terraform-bot-account"
         },
         "Action": [
            "s3:GetBucketLocation",
            "s3:ListBucket"
         ],
         "Resource": [
            "arn:aws:s3:::pegski-tfstate"
         ]
      },
      {
         "Sid": "statement2",
         "Effect": "Allow",
         "Principal": {
            "AWS": "arn:aws:iam::<insert-arn-number>:user/terraform-bot-account"
         },
         "Action": [
             "s3:GetObject"
         ],
         "Resource": [
            "arn:aws:s3:::pegski-tfstate/*"
         ]
      }
   ]
}

and the inline policy for the aim user terraform-bot-account

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "PermissionForObjectOperations",
            "Effect": "Allow",
            "Action": [
                "s3:PutObject"
            ],
            "Resource": [
                "arn:aws:s3:::pegski-tfstate/*"
            ]
        }
    ]
}

deploy the infrastructure

Ensure the environment variables are loaded in your terminal sessions and that you are in the terraform directory. Then simply type:

make apply

Warning: creating the redis cluster can take a significant amount of time (> 5 minutes). Just be patient and let it finish.

useful commands for testing the infrastructure

You can use ansible with the terraform.py script from cisco to quickly ping all created instances:

ansible -i bin/terraform.py/terraform.py -m ping all

dockernode-02 | SUCCESS => {
    "changed": false, 
    "ping": "pong"
}
haproxy-01 | SUCCESS => {
    "changed": false, 
    "ping": "pong"
}
dockernode-01 | SUCCESS => {
    "changed": false, 
    "ping": "pong"
}
mongodbnode-01 | SUCCESS => {
    "changed": false, 
    "ping": "pong"
}

destroying the infrastructure

(be carefull ;) If you want to destroy the infrasctructure, use:

make destroy

The command will wipe the complete infrastructure. You can also be more specific. If you want for example to create new dockerhosts, you can use:

terraform destroy --target module.mongodbnodes.aws_instance.dockernodes

provisioning of the hosts

All the hosts can be provisioned using:

ansible-playbook -i terraform/bin/terraform.py/terraform.py ansible/pegski.yml

There also the possibility to be more specific:

 ansible-playbook -i terraform/bin/terraform.py/terraform.py ansible/pegski.yml --limit "role=dockernodes"

left todo:

  • fix security groups, permissions are to wide now while dockernodes cannot access mongodb
  • setup memcached
  • refactor dockerfiles
    • files should be loaded in docker container
    • switch from nginx to apache to allow env variables
    • adjust SF3 configuration to use env variables for configuration
    • update SF3 mongodb settings to use dedicated mongodb instance
  • setup jenkins to listen for github webhook to run tests and deploy
  • add basic monitoring