RFC: SOE Bundle Model

[parmstro]

Summary

This RFC proposes a bundle model for Red Hat Satellite Standard Operating Environment (SOE) configuration — a self-contained YAML unit that groups all Satellite objects required to deliver a single SOE (content, templates, lifecycle objects, and hostgroups) — and a priority-weighted dynamic snippet ordering convention that addresses a longstanding limitation in Satellite kickstart template composition.

Problem

Large-scale Satellite deployments accumulate SOE configuration across monolithic variable files that grow to tens of thousands of lines. This increases explanation cost, configuration cost, debugging cost, and makes the configuration brittle and difficult for customers to extend.

Proposed Model

• SOEs are expressed as self-contained bundle objects that aggregate to flat execution lists, preserving the existing Satellite phase-ordered execution architecture unchanged.
• Kickstart snippets are included dynamically via a host parameter (foreman_extra_snippets) using a priority-weighted ordering convention with documented phase ranges — no static snippet calls in base templates, no modification of built-in Satellite snippets required.
• The snippet ordering convention is proposed as a candidate for upstream adoption in the Foreman community.

Full Document

The complete design document — including the full SOE dependency chain, phase structure analysis, alternatives considered (including a systemd dependency graph model and why it was rejected), ERB implementation, and a staged community proposal — is available on the schema branch:

schema/soe_bundle_model.md

Call for Commentary

This is an early draft aimed at the RHIS community first. Specific questions for discussion:

1. Does the bundle declaration structure map cleanly onto your SOE definitions?
2. Are the priority phase ranges appropriate for the snippet ordering use cases you encounter?
3. Are there cross-layer dependencies or Satellite object types not covered by the model?
4. Would you use this convention independently of the broader bundle model?

All feedback welcome — inline comments, alternative approaches, and field experience with large Satellite configurations especially valued.