Fluent Commerce Logo
Docs

CI/CD Verification and Promotion

Essential knowledgeCore knowledge to understand Fluent Commerce

Changed on:

9 Nov 2026

Overview

[Warning: empty required content area]

Key points

[Warning: empty required content area]

A green pipeline is not proof of a deploy

A deploy step can succeed without changing the account. Common causes:
  • the CLI logged an error and exited 0 (see Catching errors),
  • the rules JAR version differed from `module.json`, so the CLI skipped the upload,
  • `--skip-if-installed` skipped a plugin whose version had been reused for different code,
  • the install targeted a different retailer or account than intended.
After every deploy, read back what is active and compare it with what the pipeline built.The module log shows that an install happened. For stronger evidence, also check that:
  • the active version of each workflow the release changed has the content you expect (download it with the CLI and diff it against the repository),
  • the rule plugin bundle with the expected version is active,
  • a quick smoke scenario passes.

Drift

Drift happens when someone changes an environment outside the pipeline, for example by editing a workflow in the modeller or changing a setting in the admin UI. The next deploy then overwrites that change without anyone noticing, or the change survives and the environment no longer matches any commit.
  • Detect drift regularly. A scheduled pipeline can download the active workflows and settings and diff them against the repository, failing on any difference.
  • Check for drift before a deploy too. That way an out-of-band hotfix is found and committed rather than overwritten.
  • Compare environments before promoting. A diff of workflow, module and setting versions between dev and UAT shows exactly what a promotion will change.
When you find drift, bring the change into the repository through a normal PR. Never accept a state on an account that has no matching commit.

Promotion

Promote the artifact that was proven, not a rebuild.
  • The `main` build publishes the module archive to an artifact store, either Bitbucket Downloads or your own repository, keyed by version.
  • The release pipeline downloads that exact version, checks its checksum against the one recorded at build time, and installs it.
  • Each deploy runs the same post-deploy verification as dev.
Put a manual approval in front of every deploy above dev (`trigger: manual` on the stage, and restrict who can run it in Bitbucket's deployment permissions). Agree with the client up front who approves each environment.

Rollback

Roll back by redeploying the previous release: re-run the release pipeline for the earlier `release/<tag>` branch. This works because every release is a tagged, reproducible build. Two caveats:
  • Installing workflows creates new workflow versions. Entities already in flight continue on the version they started with, so a rollback does not undo behaviour for orders that were created in the meantime.
  • Settings and reference data changes are not versioned the way workflows are. Make sure a release that changes them can be reversed by a later release, not only by hand.