caktus.k8s-hosting-services

An Ansible role to allow Caktus Kubernetes projects to take advantage of common Caktus-provided services. These include:

• backing up the Postgres database to the backup AWS S3 bucket.
• backing up the environment S3 bucket to the backup AWS S3 bucket.
• More coming soon! (Papertrail, NewRelic infra, etc)

License

This Ansible role is released under the BSD License. See the LICENSE file for more details.

Development sponsored by Caktus Consulting Group, LLC.

How do I add this role to my project

1. Add to your requirements.yaml:

   ---
   - src: https://github.com/caktus/ansible-role-k8s-hosting-services
     name: caktus.k8s-hosting-services
     version: v0.7.0

   If you're using invoke-kubesae, run: inv deploy.install to install the Ansible requirements.

2. Create a deploy/deploy-hosting-services.yaml file which you'll run once to set up the backup process (and then infrequently whenever you'd like to make changes to it). Put this inside of it:

   - name: caktus hosting services
     hosts: production
     vars:
       ansible_connection: local
       ansible_python_interpreter: "{{ ansible_playbook_python }}"
     gather_facts: false
     roles:
       - role: caktus.k8s-hosting-services  # backups only
     # Install configured monitoring tools (optional):
     tasks:
        - import_role:
            name: caktus.k8s-hosting-services
            tasks_from: monitoring

3. Ensure the production host is in your deploy/inventory file:

   ...
   [k8s]
   staging
   production

4. Method 1: Client AWS account and S3 bucket (preferred)

   Clients typically prefer backups to be stored within their own AWS account. This doesn't necessarily need to be the same account as the production K8s cluster, but for simplicity it is usually set up this way.

   1. Create new S3 bucket to store backups, e.g. project-production-backups
   2. Create new caktus-backup IAM user in client AWS account and create a new access key. Save the AWS_SECRET_ACCESS_KEY for use below.
   3. Attach a restricted IAM policy (remember to fill in the real bucket name) to the user:

   {
      "Version": "2012-10-17",
      "Statement": [
         {
               "Sid": "ListObjectsInBucket",
               "Effect": "Allow",
               "Action": [
                  "s3:ListBucket"
               ],
               "Resource": [
                  "arn:aws:s3:::project-production-backups"
               ]
         },
         {
               "Sid": "ObjectActions",
               "Effect": "Allow",
               "Action": "s3:PutObject",
               "Resource": [
                  "arn:aws:s3:::project-production-backups/*"
               ]
         }
      ]
   }

5. Method 2: Caktus Hosting Services AWS Account and caktus-hosting-services S3 bucket

   The AWS_SECRET_ACCESS_KEY is in the LastPass shared entry: "Caktus Website Hosting Services k8s secrets".

6. You'll need to encrypt at least 2 variables for this role to work:

   1. Your production DATABASE_URL.
   2. The AWS_SECRET_ACCESS_KEY for the backup IAM user.

   The first should be obtained from your project's configuration and AWS_SECRET_ACCESS_KEY is found above.

   Use ansible-vault to encrypt them:

   ansible-vault encrypt_string SECRET-VALUE-HERE

7. Once you have those encrypted values, add the following to deploy/group_vars/production.yaml:

   k8s_hosting_services_project_name: "your-project-name"
   k8s_hosting_services_healthcheck_url: "<... project healthcheck url ...>"
   k8s_hosting_services_database_url: "<... secret from ansible-vault output ...>"
   k8s_hosting_services_backup_base_bucket: "<... project backup S3 bucket ...>" OR "caktus-hosting-services"
   k8s_hosting_services_aws_region: "<... bucket region ...>"
   k8s_hosting_services_aws_access_key: "<... project iam user access key ...>"
   k8s_hosting_services_aws_secret_access_key: "<... secret from ansible-vault output ...>"

   • You should generate a new healtcheck url at healtchecks.io and provide that to the developer for their k8s_hosting_services_healthcheck_url variable.
   • k8s_hosting_services_project_name will be the directory in S3 under which these backups will be stored.

8. By default, this role will run backups on a daily, weekly, monthly and yearly schedule. If you don't need all of those, or if you need a custom schedule, then override k8s_hosting_services_cron_schedules. Ideally, you should do this before your run this role for the first time, because this role does not delete existing schedules. If you do need to delete a schedule, you can do it manually using k8s commands. Here is an example of customizing your schedule to remove daily backups and add a rule to backup every 2 hours:

   k8s_hosting_services_cron_schedules:
     - label: weekly
       schedule: "@weekly"
     - label: monthly
       schedule: "@monthly"
     - label: yearly
       schedule: "@yearly"
     - label: every2hours
       schedule: "0 */2 * * *"

9. Review defaults/main.yml in this repo to see other variables that you can override.

10. See the next section for the commands to deploy this role to your cluster.

How do I deploy this role to my cluster

Once you have configured the role as described above (or any time you need to make a change to the configuration), you can deploy this to your kubernetes cluster.

• Using invoke-kubesae:

  inv deploy.playbook -n deploy-hosting-services.yaml

• Without invoke-kubesae:

  cd deploy/
  ansible-playbook deploy-hosting-services.yaml -vv

Run a database backup job manually

By default, the shortest backup cronjob interval is daily, which means that the first job won't run for roughly 24 hours after the initial setup. This makes it difficult to test if backups are configured correctly.

You can manually invoke a job like so:

kubectl create job --from=cronjob/backup-job-daily test-job-0001 -n hosting-services

Papertrail

Add the following for each cluster to monitor:

k8s_papertrail_logspout_destination: syslog+tls://YYYYY.papertrailapp.com:NNNNN
k8s_papertrail_logspout_syslog_hostname: "{{ k8s_cluster_name }}"

New Relic Infrastructure

New Relic's Helm Charts are used to install New Relic Infrastructure monitoring.

Add the following for each cluster to monitor:

# https://github.com/newrelic/helm-charts/releases
k8s_newrelic_chart_version: "2.22.3"
k8s_newrelic_license_key: !vault...

Database restore

Create a playbook like so:

# file: db-restore.yaml
- hosts: k8s
tasks:
- import_role:
      name: caktus.k8s-hosting-services
      tasks_from: restore

See defaults/main.yml for variables to use.

Maintainer information

If you are working on the role itself (rather than just using the role), make sure to set up a Python 3 virtualenv and then set up pre-commit:

pip install -Ur requirements.txt
pre-commit install  # <- only needs to be done once

The pre-commit tasks will run on each commit locally, and will run in Github Actions for each pull request.