Raft in Kubernetes and Swarm Pt. 3: Backup the Raft

Pt.1 and 2 covered the Raft algorithm and cluster design, and today I’ll cover backing up your cluster database.

Since Swarm has a built-in Raft database (called Raft logs), this is cut and dry, you just need to backup everything in /var/lib/docker/swarm on a Manager node.

The negatives are that 1. That Manager’s Docker Engine should be stopped during file backups, which means you need to take steps to ensure the Swarm still has quorum while backups happen, and 2. There is no built-in way to automate this. You’ll need to create or find scripts/tools that do this outside a Docker container on a specific server, and then move those backups off-disk. Bummer.

Since etcd can be installed in all sorts of ways (hosted, in cluster containers, outside the cluster, or directly on hosts), this makes backups more complex and varied. The good news is Kubernetes has tools for making this easier.

What a fun show, talking about where Jenkins X, how it’s different from Jenkins (hint, it’s totally unrelated), and how to use it for proper Kubernetes GitOps (DevOps-style cluster updates using git).

Thanks, talk to ya next week,

–Bret

Weekly YouTube Live: bretfisher.com/youtube

Course Coupons: bretfisher.com/courses

Podcast: bretfisher.com/podcast

Twitter: twitter.com/bretfisher