Passarelli
06/08/2026, 2:21 PMkapa.ai
06/08/2026, 2:21 PMskip_full_check_for_parquet: true has been introduced (in connector version 4.15.3) that skips the full record-parse step during CHECK for Parquet streams. This prevents OOM errors on large Parquet datasets. Note the trade-off: with this flag enabled, schema mismatches or corrupt files will only surface at sync time rather than during CHECK. [S3 OOM PR]
Recommended Steps 1. Check your Parquet file sizes â if they are large (hundreds of MB or more), the OOM issue is likely the cause.
2. Upgrade to S3 connector version 4.15.3 or later and enable skip_full_check_for_parquet: true on the affected stream if OOM is the issue.
3. Increase ingress timeout on your Kubernetes ingress controller if the timeout is hitting before the check completes.
4. Allocate more memory to the connector job if possible.
Could you share more details about your deployment type (Docker/Kubernetes) and the approximate size of your Parquet files? That would help narrow down the root cause further. đ 0 button đ 0 button âšī¸ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.