Skip to main content

Troubleshooting

Common Issues​

1. PostgreSQL Connection Failed​

  • Ensure PostgreSQL is running in codemie namespace
  • Check PostgreSQL pod name and connectivity
  • Verify PostgreSQL service name

2. Helm Deployment Failed​

  • Review values.yaml configuration
  • Check resource availability in cluster
  • Verify Helm chart version compatibility
  • Ensure domain placeholders are replaced

3. Errors in LangFuse components​

  • Check that resource limits are appropriate
  • Follow system requirements outlined in this document or in the official guide

4. ClickHouse runs out of resources​

  • Make sure that ClickHouse pods have at least 8 Gb of memory
  • Follow LangFuse scaling recommendations

Recovery​

If deployment fails:

1. Clean up resources​

# Remove namespace
kubectl delete namespace langfuse

# Delete langfuse-integration secret
kubectl -n codemie delete secret langfuse-integration

# Reset postgresql database instance
# Connect to the "postgres" database first, if not already connected
psql -U postgres

# -- Revoke schema and database privileges
REVOKE ALL ON SCHEMA public FROM langfuse_admin;
REVOKE ALL ON DATABASE postgres FROM langfuse_admin;

# -- Drop the user
DROP USER langfuse_admin;

# -- Drop the database
DROP DATABASE IF EXISTS postgres_langfuse;

2. Fix configuration issues​

Review and update your configuration files based on the error messages.

3. Re-run deployment​

Choose one of the following:

  • Automated: ./deploy-langfuse.sh
  • Manual: Follow manual steps again