Backup and recovery

Backing up enterprise applications requires much more than snapshotting a few persistent volumes and backing up a dozen etcd objects. That approach might work fine for a simple deployment running in a single namespace. Enterprise applications go well beyond that. They consist of operators, operator groups, custom resources, catalogs, subscriptions, and they often span multiple namespaces.

Online backups are critical for enterprise applications. You can’t afford to schedule your backups around maintenance windows. And yet you also need application consistent backups, ensuring that memory buffers are drained to disk before snapshots are run so that no data is lost.

Fusion provides a solution for these complexities called recipes. A recipe is a custom resource that instructs Fusion on how to orchestrate both the backup and the recovery of an application. Recipes ensure that consistent online backups of applications are available, and also ensure that applications come online cleanly during a restore.

Learn more about considerations for backing up and restoring enterprise applications

A recipe orchestrates backup by identifying groups of resources that need to be backed up. This could be groups of etcd resources, or groups of persistent volumes that need to be snapshotted at the same time. A recipe also includes a set of hooks. Hooks are commands that are run against pods to do things like quiesce and drain workloads, or to resume workloads after a snapshot.

The backbone of a recipe is the workflow. The workflow instructs Fusion on the order that groups should be backed up, and when hooks should be run. This all comes together to achieve an application consistent backup while the application remains online.

Learn more about recipes

There’s an industry joke that backing up an application is the easy part; it’s restoring it into a working state that’s the tricky part. Fusion uses recipes to orchestrate the recovery of an application so that it comes back online cleanly. Similar to what’s described above in the section on backup, restoring using a recipe hinges on a workflow, which consists of a sequence of groups of resources that are restored and hooks that are run.

For example, you can’t start restoring subscriptions until catalogs are ready. And you can’t start restoring custom resources until subscriptions are ready and custom resource definitions are in the cluster. As your workload comes back up, you may need to trigger a command to instruct it that it is recovering from a failure. Fusion orchestrates all of this using recipes.

Fusion has built-in recipes for working with IBM Cloud Paks - IBM’s enterprise ready, containerized software solutions for applications, integration, and more. For example, Fusion can automatically backup IBM Cloud Pak for Data instances, and restore them to an alternative OpenShift cluster. This is all achieved using built in recipes - no assembly required.

Fusion’s open source repository hosts recipes for popular applications such as DB2, Redis, MongoDB, and Postgres. Join the community and help contribute to the library of recipes for enterprise applications.

Go to the open source recipe repository