Management Packs
For whole-store import, service selection, complete configuration files, and durable operation recovery, use the runtime mpack store guides. They describe the AMBARI-26663 development follow-up.
The command-line workflow below describes the older setup-time mechanism at its pinned source revision. Its staging/restart behavior must not be assumed for every runtime store operation. Use the manifest and tooling expected by the chosen mechanism; sharing the name mpack.json does not establish archive compatibility.
An Ambari management pack is an archive of stack, service, extension, or view artifacts plus mpack.json metadata. The Server setupMpacks.py implementation expands the archive, reads its metadata, validates prerequisites, stages the pack, and creates the resources used by the stack loader.
Metadata
mpack.json identifies the pack with a name, version, and description. Its prerequisites can constrain Ambari versions and installed stack versions. Its artifacts describe the source directories and artifact types, such as stack definitions, service definitions, extension definitions, and stack add-on service definitions. Invalid or missing metadata prevents installation.
Packaging and Dependencies
Package only the definitions and files required by the artifact catalog. A full stack pack can provide stack versions and common service definitions; an add-on pack can provide a service and its applicability to an installed stack. Keep package scripts, configuration, repositories, and service metadata in the paths named by each artifact. The management-pack mechanism is independent of the retired AMS/Ganglia packaging model.
Install, Upgrade, and Purge
The installer validates prerequisites and can run pack hooks around installation or upgrade. Purging Stack definitions is restricted to packs that contain Stack artifacts, preventing an add-on pack from accidentally removing an installed Stack. After staging, Ambari updates resource links and explicitly requests an Ambari Server restart; staging alone does not reload the running Server's Stack model.
In a planned maintenance window, install a reviewed pack and restart Server:
sudo ambari-server install-mpack --mpack=/path/to/reviewed-management-pack.tar.gz
sudo ambari-server restart
Confirm the newly loaded Stack/service metadata and a controlled service operation afterward. Do not add purge options to a normal add-on installation or treat the archive's successful extraction as completed runtime validation.
Verification
Inspect mpack.json, artifact paths, service descriptors, package files, configuration dependencies, and prerequisites before deployment. Test installation, upgrade, rollback or uninstall behavior, and service checks on the target stack. The implementation is in setupMpacks.py.