Odoo hosting.

Failover / Fallback

In the case of an unexpected issue at the hardware/VM vendor or their network, your Odoo instance can self heal or recover from the issue and becomes back available in a minimal downtime if there is a failover or fallback strategy implemented.

With a failover/fallback strategy there is a minimal intervention required to make sure your base cluster is always in the correct dimension to handle application loads. 

With Kubernetes, it's even possible to achieve high availability so you don't experience any downtime at all and it runs on auto pilot. 

If you don't value this option and don't mind interruptions, you can select no.
If you do want this feature, the minimum specification for deploying Odoo requires a cluster setup. 

Auto scaling

Auto scaling is a feature exclusively available in Kubernetes only. This feature allows your cluster to spin up and down VM's and resources based on metrics and health checks to maximize uptime and performance of your applications. 

If you run very dynamic work sets or you have a very dynamic traffic with unexpected high loads, or you often run large campaigns that cause traffic spikes, then Kubernetes might be an interesting feature for you. With a minimum and maximum set, Kubernetes will spin up servers and keep them available until the sustained load is decreasing or completely gone and then automatically removes the temporary VM's from the cluster. This way, you can guarantee a high performance solution for dynamic applications and don't waste resources on running VM's 24/7 that don't handle any load at all. 

Preview deployments

Preview deployments is a dynamic feature in combination with Github pull requests and Argocd. 
Each time you git commit and push a change to your Github repository, and trigger a pull request it will launch a staging/dev instance of your Odoo instance on a random preview URL based on the PR commit ID. These instances are ephemeral. After you test your changes and merge your pull request, Argocd will also cleanup those temporary instances. 

This feature is mostly a recommendation for companies that are in a very dynamic development situation and want a secure and automated process of developing > testing > delivery cycle without the manual hassle of creating new setups by hand.