Why Replication Alone Does Not Protect Your Data

▶ Watch (0:02)

Distributing nodes across availability zones sounds safe until the zones turn out to share physical space. Aukia cited two real incidents: one where racks assumed to be separate were stacked on top of each other and burned together, and one where a misconfigured cluster placed most nodes in a single zone. That zone went offline, the service went down for hours, and availability zone labels gave false confidence. Backups fix what replication cannot.

Backup as a Service: Shifting Ownership to Application Teams

▶ Watch (0:02)

Most backup operators before 2018 operated from the infrastructure side. Platform engineers decided what to back up and how often, without knowing each workload’s restore point or restore time objectives. K8Up flips that. The platform engineer provides the backup service. Application teams use Kubernetes CRDs to declare what data matters, how often it backs up, where it goes, and what the security requirements are. The people who know the application make the backup decisions.

How K8Up Works: restic, CRDs, and Storage Backends

▶ Watch (3:05)

K8Up wraps restic, a backup tool that predates K8Up by several years. restic encrypts every backup using a repo password pulled from a Kubernetes secret. It supports several dozen storage backends, including S3, also configured via secret. Users define a schedule directly on their namespace: nightly backup, weekly prune, monthly check. For stateful applications, backups run inside an existing pod or a new pod that streams data out. Restores and backup listings use the same CRD model, so teams restore themselves without calling the platform team.

Project Status and How to Get Involved

▶ Watch (4:38)

K8Up launched at KubeCon Barcelona 2019 and remains a CNCF sandbox project. The team is working toward incubation and needs contributors across engineering, technical writing, and product management. Maintainers Tobias, Nicholas, Lena, Sabi, Simon, and Gabrielle keep the project moving. Anyone interested can connect on Slack or meet the maintainer quorum at the project pavilion the following day, 10am to 2pm.

Notable Quotes

not a backup But the backup as a service, right? So you as a platform engineer, you provide the backup service Aarno Aukia · ▶ 02:30

we use uh a battleproven uh open source backup tool called restic um that’s been around for way longer than we have we’re just auler and the kubernetes packaging around that Aarno Aukia · ▶ 03:05

no backup makes sense without a restore Aarno Aukia · ▶ 04:27

Key Takeaways

  • Availability zone distribution failed two real operators when physical placement assumptions were wrong.
  • K8Up moves backup ownership to application teams using Kubernetes CRDs and GitOps config.
  • restic handles encryption and storage; K8Up is the Kubernetes scheduling and CRD layer on top.

About the Speaker(s)

Aarno Aukia is a co-founder and partner of VSHN - The DevOps Company in Zürich, Switzerland. He has worked on DevOps and platform engineering with Kubernetes and open-source software for over ten years, and is a maintainer of the K8Up CNCF sandbox project.