Skip to main content

Tag Database Changeset

Last updated on

The Tag Database Changeset option is available on the Apply Schema step under Harness Database DevOps. When enabled, it ensures that a post-deployment tag is recorded even when a deployment applies zero changesets, giving every deployment a stable rollback anchor regardless of whether schema changes were made.

Why this matters​

By default, Harness skips the post-tag apply when a deployment results in zero applied changesets (a no-op deployment). This means no tag row is written to DATABASECHANGELOG for that run.

In multi-service release workflows this creates a gap: if you need to roll back all services to a consistent state, a service that had a no-op deployment has no tag anchor for that release. Rolling back in sync across services becomes unreliable.

Tag Database Changeset closes this gap by writing a synthetic changeset row into DATABASECHANGELOG with the post-tag value, even when there are no real schema changes to apply.

How it works​

When Tag Database Changeset is enabled and a deployment results in zero applied changesets, Harness checks whether the post-tag already exists as the latest row in DATABASECHANGELOG:

  • Tag already exists on the latest row: Harness skips insertion and logs a notice.
  • No tag on the latest row: Harness updates that row with the tag value directly. No synthetic row is created.
  • A different tag on the latest row: Harness inserts a synthetic tag-marker row with a stable identity so rollback can detect and clean it up:
    • Changelog path: liquibase-internal
    • Author: harness
    • ID: harness_<sanitizedTag> (non-alphanumeric characters in the tag are replaced with _; for example 1.1.0-4.33 becomes harness_1_1_0_4_33)
    • The row is recorded with status EXECUTED and synced to Harness migration state and command execution history.

When a deployment applies one or more changesets, the post-tag runs normally and no synthetic row is created, regardless of whether the checkbox is enabled.

Enable Tag Database Changeset​

  1. In your pipeline, add or edit an Apply Schema step.
  2. In the Apply Schema step, under Specific Tag to Create, add the tag value you want to use for the post-deployment tag.
  3. Under Global Settings, check the option:
    • Tag Database Changeset (disabled by default).
  4. Save your changes and run the pipeline.

Tag Database Changeset Option

Rollback behavior​

When rolling back to a tag, Harness performs additional cleanup for synthetic tag-marker rows:

  • After the rollback completes, Harness removes synthetic tag-marker rows that occur after the rollback target tag.
  • The synthetic marker row for the rollback target tag itself is preserved as the rollback anchor.
  • Each removed orphan row is logged in the step output as Rolling Back Changeset: <changeset-id>.

When to use Tag Database Changeset​

Use this option when:

  • You deploy across multiple services and need every service to have a tag anchor for each release, including releases where a service had no schema changes.
  • You rely on rollback-to-tag for synchronized multi-service rollbacks.

Do not enable this option if:

  • You do not use post-tags on the Apply Schema step (the option has no effect without a configured post-tag value).
  • Your workflow treats no-op deployments as non-events and does not require a tag anchor for them.

FAQ​

Does this option affect deployments that apply changesets?​

No. When one or more changesets are applied, the post-tag runs normally and no synthetic row is created. The option only changes behavior when zero changesets are applied.

What happens if I run the same pipeline twice with no new changesets?​

On the first run, if the post-tag does not yet exist, Harness creates the synthetic row. On the second run, Harness detects the tag already exists on the latest row and skips insertion, logging a notice.

Is the synthetic row visible in the Harness changeset history?​

Yes. The synthetic tag-marker row is synced to migration state and command execution history with a Success status, so it appears alongside real changeset deployments.

Can I roll back to a synthetic tag-marker row?​

Yes. Rollback-to-tag treats the synthetic row as a valid rollback anchor. After rolling back to it, Harness cleans up any orphan synthetic marker rows that occurred after the target tag.

Does this replace the preStartTag that Harness creates before the deployment?​

No. The preStartTag (created before the Apply Schema step runs) and the post-tag (created after) serve different purposes. Tag Database Changeset only affects the post-tag behavior on no-op deployments. Go to Rollback Tags with Apply Schema to understand how preStartTag works.