What Network Access Skippr Needs for Delta Lake

June 2026

Delta Lake source access is really object-store access to a specific table URI, including the Delta log and the data files behind it.

Short Answer

Skippr needs read access to the exact Delta Lake table_uri and to the storage backend behind that URI. The Delta source docs define the connector around one table location such as s3://my-bucket/path/to/table, an optional version, and optional storage_options, so the useful network boundary is the table root plus every object read needed to open the Delta transaction log and the referenced data files.

That means Delta Lake access is not just can we see the bucket. If the runner can list a folder but cannot read _delta_log metadata or the Parquet files the log points at, the table is still unreachable in practice. A common case is a table on S3 where the bucket path resolves fine, but the runner lacks the region or credential settings required for the actual objects that make up version 5 of the table.

Why Teams Struggle with This

Delta Lake sources are easy to describe and easy to half-configure. Teams often have the table URI in hand, but the missing piece is that Skippr reads a table, not a loose folder, so access has to cover the log and the files together.

• The table_uri must point at the Delta table root, not a nearby prefix that only looks similar.
• The runner needs storage credentials that can read the Delta log and the underlying data files.
• A pinned version is useful for repeatability, but it still depends on the same object-store access.
• A reachable bucket path is not enough if backend-specific storage_options are incomplete.

How Skippr Handles It

Skippr exposes the real Delta boundary directly: one table_uri, optional storage_options, optional version, optional filter. That makes it easier to ask for the right access from a cloud platform team because the request is attached to one concrete table location instead of a generic lake permission set.

It also keeps the first test honest. If Skippr can open the Delta log, resolve the selected snapshot, and scan the rows you expect, then the network and storage path is actually ready. If it cannot, there is no value in pretending the connector is mostly connected.

• Reads Delta tables from s3://, ADLS, or local paths through one table URI.
• Uses backend-specific storage_options for the real object-store access path.
• Can pin a table version when you want a repeatable validation point.
• Keeps the source definition centered on the table boundary instead of a file-dump convention.

What the First Useful Version Looks Like

The first useful version is one Delta table URI, one small validation run, and a credential path that works from the same machine or job that will run production syncs.

If the table opens only from your laptop shell and not from the runner environment, the source is not ready yet. Fix the storage path before you widen the table set.