Area: Ingestion Issues
Sub-Area: LookML Connector
Issue
LookML ingestion jobs fail after a DataHub upgrade with SSH-related errors when the connector attempts to clone Git repositories. This occurs when the openssh-client package is missing from the executor container image, preventing Git operations that require SSH authentication.
Error Messages
ssh: not foundfatal: Could not read from remote repository.
You Might Be Asking
- Why did my LookML ingestion suddenly stop working after an upgrade?
- Is this a configuration issue with my Git repository or SSH keys?
- How can I fix SSH connectivity issues in LookML ingestion?
Solution
This is a known regression that can occur when DataHub's executor container image is updated without including the openssh-client package. The issue typically affects LookML connectors that use SSH to clone Git repositories from GitHub, GitLab, or other Git hosting services.
For DataHub Cloud Users:
- Contact DataHub Support immediately if you encounter this issue after an upgrade
- No action is required on your end - this is a platform-side fix that requires updating the executor container image
- Monitor your ingestion jobs and rerun them once the fix has been deployed to your environment
For Self-Hosted Deployments:
- Check if openssh-client is installed in your datahub-executor container:
kubectl exec -it-- which ssh - If SSH is missing, update your executor image to a version that includes openssh-client, or modify your Dockerfile to include it:
RUN apk add --no-cache openssh-client - Rebuild and redeploy the executor container
- Restart any failed LookML ingestion jobs
Verification Steps:
- After the fix is deployed, rerun your LookML ingestion jobs
- Check that the ingestion completes successfully without SSH errors
- Verify that your Looker metadata is being properly ingested
Additional Notes
This regression commonly occurs when DataHub's base container image is migrated from Ubuntu (which includes openssh-client by default) to minimal distributions like Alpine or Wolfi that don't include SSH tools by default. The LookML connector specifically requires SSH access because it uses Git over SSH to clone Looker project repositories. For DataHub Cloud customers, this is managed infrastructure and requires a platform-side fix. Self-hosted users should ensure their executor images include all necessary system dependencies. Consider setting up ingestion failure notifications under Settings → Notifications to catch similar issues early.
Related Documentation
Tags: lookml, ingestion, ssh, git, container, upgrade, regression, executor, openssh-client, deployment