Area: Ingestion Issues
Sub-Area: Snowflake / Memory Management
Issue
Following recent DataHub CLI upgrades, Snowflake ingestion jobs may fail mid-run due to out-of-memory (OOM) conditions. The ingestion process is forcibly terminated by the operating system before it can complete, causing the run to appear as failed in the DataHub UI with no partial results written. This behavior affects CLI versions prior to v1.6.0.3 and is resolved by upgrading.
Error Messages
Exit code: 247Process terminated: signal 9 (SIGKILL)peak_memory_usagein run logs matching or approaching themem_infoceiling
You Might Be Asking
- Why did my Snowflake ingestion start failing after a CLI upgrade?
- What does exit code 247 mean in DataHub ingestion?
- Why is my ingestion being killed mid-run with no clear error message?
Solution
Upgrade your DataHub CLI to version v1.6.0.3 or later. This release includes memory management improvements that prevent the ingestion process from exceeding container memory limits during Snowflake metadata extraction.
To upgrade, run:
pip install 'acryl-datahub=1.6.0.3'
After upgrading, re-trigger the failed ingestion recipe. No changes to your recipe configuration are required.
If you are running ingestion through a remote executor, contact DataHub Support to confirm your executor has been updated to a compatible version before re-running.
Additional Notes
Exit code 247 is how a -9 SIGKILL return code surfaces when interpreted as an unsigned byte (256 − 9 = 247). It is a reliable indicator that the ingestion child process was forcibly terminated by the OS due to memory exhaustion — not a recipe misconfiguration or a Snowflake-side error. You can confirm OOM as the root cause by checking the ingestion run logs for a peak_memory_usage value that matches or approaches the value shown in mem_info.
Related Documentation
Tags: snowflake, ingestion, oom, out-of-memory, exit-code-247, sigkill, cli-upgrade, memory, remote-executor, v1.6.0.3