bkatiemills / argovis_deployment

[wip] Argovis Deployment

This repo will contain notes and resource on deploying Argovis to production.

Release Process

Argovis consists of four microservices, wihch all have their own release process (except for mongo, for which we use the off-the-shelf upstream). See the latest releases of each in the following repos:

• API
  • GitHub: https://github.com/argovis/argovis_api
  • Docker Hub: https://hub.docker.com/repository/docker/argovis/api
• Redis
  • GitHub: https://github.com/argovis/argovis_redis
  • Docker Hub: https://hub.docker.com/repository/docker/argovis/redis
• React:
  • GitHub: https://github.com/argovis/react
  • Docker Hub: https://hub.docker.com/repository/docker/argovis/react

Deploy to Kube

Kubernetes is Argovis' preferred deployment target, when a managed cluster is available. To this end, we provide a Helm chart in this repo. This chart assumes there exists the following objects in the cluster:

• a PVC which contains everything from /data/db in the mongo container.
• a secret called sendgrid-api-token, which contains a key SENDGRID_API_TOKEN, with a value that will be used against the Sendgrid API.
• an API token active in mongodb's user table with the kv "key": "guest"

When any of the underlying base images change, or anything about the orchestration structure changes, follow the procedures below to increment the release on the OKD cluster:

• In argovis/values.yaml, update any image tags as necessary.
• In argovis/Chart.yaml, increment the version and appVersion values. Keep these the same, and use semver.
• Log into the OKD cluster from the command line by doing oc login, and following the link it provides.
• Update the application: helm upgrade argovis ./argovis
• Make a new release in this repo tagged the same as the chart version.

Some helpful things to know about application management that you may occassionally need:

• Delete / recreate application: helm uninstall argovis / helm install argovis ./argovis
• Undo last update: helm rollback argovis

Deploy to Swarm

In the event that Argovis must be deployed to an environment where a managed Kubernetes service isn't available, we also provide a script to deploy services and containers to Docker Swarm. Read swarm-deploy.sh and make sure image tags and environment variables are up to date, and deploy to a host with sufficient resources and Docker installed with bash swarm-deploy.sh.