We are excited to announce the release of Ark v0.9.0! This release brings two major new features: integration with restic to back up almost every type of Kubernetes volume, and initial support for exporting metrics in the Prometheus data format. There is also a critical bug fix to avoid potential backup/restore data corruption, so we encourage you to upgrade.
Volume snapshots with restic
With Ark v0.9.0, it’s now possible to snapshot almost every type of Kubernetes volume (hostPath is not supported). This feature complements Ark’s ability to snapshot disks from AWS, Google, and Azure (as well as any custom plugin integrations, such as PortWorx). You could use native snapshotting for these types of disks and restic snapshots for everything else, or you could use restic for everything. Ark’s restic integration is a good option if your type of PersistentVolume doesn’t currently have a plugin for Ark, or if it doesn’t offer a native snapshot concept, such as emptyDir and NFS.
For more details, including setup instructions, see our restic documentation and our initial annoucement.
Prometheus metrics
We’d like to offer a huge thanks to community contributor Ashish Amarnath for providing the initial support for exposing Prometheus metrics. Ark v0.9.0 includes the following metrics:
- total # of backup attempts
- # of successful backups
- # of failed backups
- backup duration histogram
- backup tarball size
All of these metrics are grouped by schedule.
We are working to add additional metrics for backups as well as restores. If you’re interested in discussing what kinds of metrics you’d like us to add, please feel free to share your thoughts on our open issue.
Additional v0.9.0 highlights
Ark now restores any image pull secrets or other secrets that you added to the default service account.
Ark also automatically backs up all cluster roles and cluster role bindings that reference a given service account. If you’re backing up one or more specific namespaces and not including all cluster-scoped resources, this feature ensures your backup isn’t missing any relevant cluster-scoped RBAC resources.
This release includes several other improvements and bug fixes:
- Ark no longer tries to restore completed jobs, completed pods, or mirror pods
- Ark no longer backs up terminating resources
- Ark no longer backs up the same replica set or daemon set twice
- Ark no longer restores a PV with a reclaim policy of “delete” when there is no associated snapshot
- Ark works more smoothly with OpenShift
- We have improved our error handling, especially when backing up pods, their PVCs, and the associated PVs; as well as marking a backup as “failed” when uploading to Google Cloud Storage fails
- All logging from the Ark server now writes to stdout instead of stderr
See the release notes for full details on all improvements and bug fixes.
What’s next?
We are actively working on designing our next big feature — replication. This will ensure that your backed up Kubernetes resources and your persistent data are available in multiple locations, avoiding single points of failure.
We are also continuing to plot the roadmap to Ark 1.0. We’ll be discussing our plans with the community in the following weeks and encourage you to join the our Google group and Slack channel.
Finally, if you’re interested in contributing, you’ll find several GitHub issues labeled as Good First Issue and Help Wanted. Take a look — we would welcome your participation.