Decaf Blog Data pipelines, open-source tooling, reverse engineering, and systems notes

ingestr v1.0.66 - OneLake Directory Skipping

3 min read

bruin-data/ingestr v1.0.66 was published on July 6, 2026, with a narrow OneLake fix: Fabric managed area directories are skipped during OneLake table path handling. For data engineering teams moving data into Microsoft Fabric, that matters because managed storage areas should not be treated like ordinary table paths by a loader or sync process.

The full release notes and downloads are on the GitHub release page.

OneLake traversal now avoids Fabric managed areas

The changelog names onelake: skip fabric-managed area directories as the only code change in v1.0.66. The useful point is not the size of the diff. It is the boundary it draws between data that ingestr should inspect and storage that Fabric owns.

Fabric managed area directories are an implementation detail of the lake environment. A path walker that treats them as ordinary directories can spend work on the wrong part of the namespace. It can also expose operators to confusing table path results, because a generated or managed area may look like candidate data when the integration is only trying to resolve a table location.

For OneLake jobs, this narrows the traversal scope. The release notes do not claim a new OneLake feature, a new flag, or a schema change. They claim a skip rule. That is still relevant for ELT jobs because path discovery logic often decides what gets copied, skipped, retried, or logged.

Why this matters in Fabric pipelines

ingestr is used as a transport tool between SaaS sources, databases, warehouses, and lake targets. In that kind of pipeline, directory semantics are part of correctness. If a connector walks into system managed paths, the failure mode is not always an obvious crash. It can be noisy listing output, retries against paths that should never be touched, or extra confusion when a job has to prove which table path was selected.

The v1.0.66 change reduces that surface for OneLake. It tells the connector to treat Fabric managed area directories as outside the data path it should operate on. That is a practical fix for operators who want logs and table resolution to reflect user data, not platform housekeeping.

Small changelog, narrow blast radius

This is not a broad ingestr release. The notes list a merge entry and the OneLake skip change. There are no release notes for other connectors, CLI flags, schema behavior, or migration steps in this tag.

That narrowness is useful when evaluating an upgrade. Teams that rely on OneLake should review v1.0.66 directly. Teams not using OneLake will probably treat it as a low context maintenance update, because the public notes do not describe changes in other connectors.

This tag is focused on path filtering. Its value is not in a new source, a new destination, or a visible workflow change.

Operator notes for rollout

Before moving a scheduled job to this tag, check whether any internal runbooks or tests assert raw OneLake directory listings. A change that skips managed areas can be correct and still alter logs, counts, or snapshots that were written around the older behavior.

Use the GitHub release page as the source of truth for the exact tag. The release is not marked as a prerelease, so there is no beta or release candidate warning in the metadata. The notes do not mention migration steps or breaking changes.

If your pipeline includes OneLake table path checks, make the first run observable. Compare the selected table path, the directory list in logs if you keep one, and the final loaded row count. The expected result is not new data shape. It is less attention paid to Fabric managed paths.

Where to get it