Guidelines to use an Air-gapped Repository

Describes the guidelines to use for proper functioning of an air-gapped repository.

Follow the guidelines mentioned below for the proper functioning of the air-gapped repository that is you have created to install or update the Data Fabric packages in your enterprise environment.

  • Always reuse the same --data-dir for cumulative updates.

  • Use --local when the canonical image has already been loaded on an isolated host and should not be pulled.

  • Do not manually edit the active repository symlink.

  • Do not manually modify files inside the active publication.

  • Do not manually delete publication directories while synchronization is running.

  • Do not place credentials inside the persistent data directory.

  • Prefer --repo-config-file for authenticated repositories.

  • Do not use docker cp to move repository content between helper and serving containers; both use the same persistent host mount.

  • Expect one active physical publication in steady state; old full repository publications are not retained for rollback.

  • Reserve enough temporary free space for the active publication plus the candidate publication during update.

  • Use remove when only the serving container should be deleted.

  • Use purge-data --yes only for permanent cleanup.

  • Serving-only lifecycle commands are offline operations and do not contact the upstream package repository.

  • Restrict host access while synchronization is running because the current mapr-setup.sh may expose parsed credentials in child-process arguments.

For independent updates, specify only the components to synchronize. Reuse the managed data root and the exact container name; omitted components remain in the cumulative publication. Include installer during initial preparation when the installation workflow needs the signing keys from its archive. The existing pub/ content is preserved when installer packages are installed. Update to the installer packages may refresh the pub/ content .