Skip to content

Latest commit

 

History

History
100 lines (71 loc) · 3.46 KB

DEPLOY.md

File metadata and controls

100 lines (71 loc) · 3.46 KB
Raw

Deploying vpc-router in production

Installing and running the code

Option 1: Installation via pip

This is the simplest way to install vpc-router:

pip install vpcrouter

Option 2: Installation via setup.py

After downloading the code, run the setup.py script, which performs a proper install and puts vpcrouter in your path:

$ git clone [email protected]:paninetworks/vpc-router.git
$ cd vpc-router
$ python setup.py install
$ vpcrouter .......

Option 3: Running in a container

Pre-built containers for the vpc-router are provided on Quay.io. You can download the latest version:

$ docker pull quay.io/romana/vpcrouter

If you prefer to build your own container, a Dockerfile has been provided with the software. To build your own container, use:

$ docker build -t yourname/vpcrouter:1.3.2 .

A few command line arguments are set by default in the container, while others still need to be specified when you run the container. Specifically, the mode as well as any parameters this mode may require.

For example, to use the configfile mode, run the container with these options, which provide a mapping for the directory containing the config file:

$ docker run -v <host-dir-with-conf-file>:/conf \
     quay.io/romana/vpcrouter -m configfile -f /conf/<conffile-name>

To run it in HTTP mode, use the following, which exposes the port 33289 on which vpc-router listens for requests. Also, a different listen address needs to be specified:

$ docker run -p 33289:33289 quay.io/romana/vpcrouter -m http -a 0.0.0.0

Note that in the container, vpc-router logs to stdout, which tends to be the preferred log destination in containerized environments.

Running on an EC2 instance in the VPC

Note: The following information is independent of how vpc-router is installed or run.

IAM policies

The most common mode of operation is to run vpc-router on an EC2 instance within the VPC for which routing tables need to be controlled. This has the advantage that it can auto-detect the AWS region and VPC (the -r and -v parameters). It also means that typically it has access to the various instances for health check purposes.

To modify VPC routes, it requires permissions to access the AWS APIs. Typically this is done by assigning IAM roles to the instance(s) on which it runs. A suitable IAM policy for the vpc-router may look like this:

{
    "Statement": [
        {
            "Action": [
                "ec2:*Vpc*",
                "ec2:*Route*",
                "ec2:Describe*"
            ],
            "Resource": [
                "*"
            ],
            "Effect": "Allow"
        }
    ]
}

Security rules for health checks

In its current form, vpc-router by default performs active health checks on all the instances listed in the route spec, by sending ICMP Echo ("ping") requests at regular intervals (--health icmpecho option). Please make sure that your security rules allow for ICMP Echo requests and responses to be exchanged.

If you use different health monitoring plugins, please adjust your security rules accordingly. For example for the 'tcp' plugin (--health tcp option), please make sure you open the port to which it attempts to connect (--tcp_check_port option).

If you use the multi plugin (--health multi) the security rules need to be configured so that all specified sub plugins (--multi_plugins ... option) can perform their security checks.