ingestr v1.1.4 - CDC Namespace Checks and URI Redaction

ingestr v1.1.4 was published on July 20, 2026. Its main operator facing change is earlier validation for CDC runs that require a destination namespace, so invalid multiple table jobs now fail fast.

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

This release changes how ingestr reports namespace configuration across two CDC execution modes. A multiple table CDC pipeline now fails fast when it has no destination namespace. That moves an invalid definition to an explicit pipeline error instead of leaving destination setup ambiguous later in the run.

The change matters for scheduled ingestion because a missing namespace is a configuration defect, not a transient source failure. Operators can classify it without waiting for retries or inspecting destination state. Existing multiple table CDC jobs should be checked for an explicit namespace before the new version is rolled out.

Single table CDC gets a warning for the inverse case. When dest_schema is supplied even though that mode ignores it, ingestr now says so. The setting still has no effect, but the warning removes uncertainty from reviews and incident work. Together, the error and warning make the accepted namespace contract clearer without changing the destination model described in the notes.

URI parsing errors no longer include the full URI. This is a narrow logging fix, but it has practical value because connection URIs can contain user names, passwords, tokens, host names, and query values. A malformed value should not copy all of that material into task logs merely to explain that parsing failed.

The scope is worth keeping precise. The release does not claim a new secret store, a general log scrubber, or automatic rotation. It changes parse error output so the complete URI is not leaked through that path. Operators should still review access to archived logs and rotate credentials if earlier errors exposed them.

This also makes failures safer to attach to tickets or send through centralized observability systems. The useful signal is the parse failure. The complete connection string is rarely needed in those downstream systems.

The connector path that fetches IDs now retrieves them in chunks. The release notes do not publish a chunk size, request count, or benchmark, so there is no basis for a specific throughput claim. The operational effect to watch is source request behavior: chunking should change how large ID sets are divided across calls, and rate limits remain the relevant constraint. Metrics for call volume and source latency should expose regressions during a staged rollout.

The changelog also records a new Payrails source and base64 support for certificates. The certificate change is useful where pipeline configuration moves through systems that handle text values more reliably than multiline certificate data. The notes do not name a configuration key or describe compatibility behavior, so deployments should validate their exact certificate input in a staging run.