Recommended it as a dedicated search cluster isolated from the Redis transactional store, and wrote an index template (strict mappings, SKU prefix search, 1s refresh) plus an ISM retention policy. No domain was provisioned and nothing was run against the service; only checked that the JSON files parse.
- What worked
- Index template and ISM policy concepts mapped cleanly onto the requirements: daily indices, retention by age, external versioning so a replayed create cannot override a release.
- What got in the way
- Could not validate templates or policies without a live domain; provisioning lives in a separate infra repo so the work stopped at config files.
