ingestr v1.1.5 was published on July 20, 2026. The main operator visible change is primary key integrity enforcement during Iceberg writes, backed by narrower fixes for MySQL CDC recovery and Snowflake schema evolution.
The full release notes and downloads are on the GitHub release page.
Iceberg writes enforce primary key integrity ¶
This release moves Iceberg primary key checks into the write path. That is the right boundary for detecting invalid records because a completed ingestion job should not leave a table with state that contradicts its declared key.
The practical consequence is stricter failure behavior. Loads that previously passed despite a primary key integrity problem may now stop during the write. That makes the error visible to the pipeline, but it can also turn latent source quality defects into failed jobs after an upgrade. Operators should test representative duplicates, null key values, retries, and merge workloads before promoting the release.
The notes do not define the exact validation algorithm, its memory cost, or its behavior for every Iceberg write mode. They also provide no throughput measurements. Large tables therefore need a normal staging benchmark, with attention to write duration and peak memory as well as correctness.
MySQL CDC recovery and Snowflake evolution ¶
MySQL CDC receives a correctness and recovery hardening change. The release notes do not identify the affected failure sequence or any checkpoint format change, so the safest validation is operational: start from a snapshot, enter CDC, interrupt the process, then resume and compare keys and row counts at the destination. Schema changes during that sequence deserve a separate test.
This is narrower than a new CDC feature. It addresses failure handling, where duplicate events, skipped events, or an unusable resume state can undermine an otherwise healthy pipeline. Teams already running MySQL CDC should keep their current reconciliation checks in place because the notes make no claim that every recovery case is covered.
Snowflake writes also gain handling for NOT NULL columns during schema evolution. That matters when a source schema changes while the destination still enforces a required column. The note does not document whether the fix changes generated DDL, column mapping, or load ordering. Test additions and type changes against a copy of the production schema rather than assuming all evolution paths now behave the same way.
Connector metadata and spreadsheet dependency fixes ¶
The Facebook Insights connector adjusts DLT load ID nullability. This is a metadata correctness change rather than a new extraction capability. Pipelines that persist or inspect DLT load metadata should verify how null values reach downstream models and assertions. No configuration or migration step is listed.
The bundled Excel processing dependency moves to excelize/v2 version v2.11.0 to fix a GHSA denial of service issue. The release notes do not include an advisory identifier or exploit conditions, but the dependency update still matters for jobs that ingest spreadsheets from outside a trusted boundary. Those operators should include malformed and unusually large workbooks in acceptance tests, then confirm that normal files retain their existing types and values.
These two items are maintenance changes. Neither expands connector scope, and the notes provide no new flags or settings for operators to adopt.