auto-cluster

README

Auto Cluster

Automatically creates a new cluster when the current one is reaching the end of its 48 hour lifespan.

Table Of Contents

• Overview
• Run
• Access Clusters
• Container

Overview

Ensures no clusters which are getting too old. If any clusters are in danger of being deleted the following steps are taken:

• Provision new cluster with the OpenShift installer tool
• Post new cluster credentials to the Slack
• Install Helm chart on new cluster
• Point DNS to new cluster

This tool is tailored for the use case of the KScout team. As such, several assumptions are made:

• DNS hosted on Cloudflare
• All applications deployed in the same namespace

Run

AWS Credentials

AWS credentials must be provided.

If you have a ~/.aws/credentials file and the credentials you wish to use are the default profile you do not have to do anything. If you do not want to use the default profile set AWS_PROFILE.

If you do not have a ~/.aws/credentials file set AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.

Configuration File

A configuration file is required. Modify the following configuration file with your information. Save as a .toml file and place in the repository root.

The auto cluster loads TOML files as configuration from the /etc/auto-cluster directory and the working directory.

[Cluster]
# Prefix to add to name when searching for / creating new clusters
NamePrefix = "NAME PREFIX"

# Oldest a cluster can be before it will be replaced
OldestAge = 42 # hours, default

# Namespace to migrate over to new development cluster
Namespace = "YOUR NAMESPACE"

[Cloudflare]
Email = "CLOUDFLARE EMAIL"
APIKey = "GLOBAL API KEY"
ZoneID = "ZONEID"

[OpenShiftInstall]
# Directory where openshift-install will store cluster details
StateStorePath = "PATH TO A DIRECTORY WHICH SCRIPT CAN WRITE TO"

[Slack]
# Slack incoming web hook used to post new cluster credentials
IncomingWebhook = "https://hooks.slack.com/services/SECRET_SLACK_INFO

[Helm]
# Git URI of repository holding Helm chart to install on new clusters
Chart = "CHART GIT URI"

Posting the new cluster credentials to Slack requires that you have an incoming web hook setup. You can set this up via the Slack API dashboard.

Dry Run

To see what the tool will do when it executes:

go run . -once -dry-run

One Time Invocation

To run the control loop once:

go run . -once

Continuous Invocation

To run every 15 minutes:

go run .

No DNS

To run the tool and ensure that no DNS changes will be made:

go run . -no-dns

Access Clusters

The auth-cluster-auth script helps provide access to temporary clusters created by the auto cluster tool.

First sync credentials down from the auto cluster instance:

./auto-cluster-auth [-n NS,-e ENV] sync

Then list available clusters:

./auto-cluster-auth [-n NS,-e ENV] ls

Finally get the copy the output of the following command and run it in your terminal:

./auto-cluster-auth [-n NS,-e ENV] env [CLUSTER_NAME]

To open the cluster's dashboard run:

./auto-cluster-auth [-n NS,-e ENV] browse [CLUSTER_NAME]

Container

The quay.io/kscout/auto-cluster:latest Docker image is available for use:

docker run \
	-it \
	--rm \
	-e AWS_ACCESS_KEY_ID=<aws access key ID> \
	-e AWS_SECRET_ACCESS_KEY=<aws secret access key> \
	-v "$PWD/config.toml:/etc/auto-cluster" \
	kscout/auto-cluster:latest

Container Development

Build and push:

make container