OpenShift 3 Backup Concept

Additionally to what’s described in the Backup Concept, the following data is backed up from OpenShift Clusters

etcd dump

Additionally to the base backup, on all master nodes which have etcd installed, a pre-backup script dumps (etcdctl backup) etcd to /var/lib/etcd.bak where it’s also backed up by burp.

OpenShift resources

A pre-backup script on each master node exports all OpenShift resources to /var/lib/openshift-backup from where it’s backed up by burp. For each resource type (for example ConfigMap, DeploymentConfig, etc.) a separate JSON file is created which contains all resources of this type.

Gluster volumes

One of the Gluster servers (configured in hiera with profile_gluster::backup) runs a pre-backup script which mounts every Gluster volume and backs up the content with burp.

This backs up Gluster volumes purely on a file level, no application-level dumps are created or similar. This means that for DBMS like PostgreSQL or MySQL, data files are backed up at runtime, which may result in corrupted files when restored.

Notable exclusions

  • The S3 storage backing the internal docker registry ISN’T backed up. Images in the internal registry are meant to be either pulled and cached from another registry, or being built reproductibly from source by a build config.