Omni Connector Troubleshooting
Use this page to find the cause of a problem with an Omni Connector. Start with the symptom, open the page that is named, and apply the fix. Connector Plane UI pages are under Settings → Connector Plane. See Connector Plane UI.
Setup and connection
| Symptom | Where to look | Cause and fix |
The shared application does not start, and its log shows Driver ... claims to not accept jdbcUrl | System Applications → OmniConnector → View status | A Shared connector's JDBC URL is wrong. Correct the field on the connector and save. See Setup. |
The shared application reports at least one connector.instances entry is required | System Applications → OmniConnector | No Shared connector exists yet. This is expected on a new plane, and Dedicated connectors do not count. Add a Shared connector. |
| The shared application stops after you added a custom property | The OmniConnector application log names the property | A custom property has a bad value. Delete that property and save. See Custom Properties. |
| Every Shared connector restarted when you saved one | n/a | Expected. Saving a Shared connector restarts the shared pod. Batch your edits, or use a Dedicated connector. See Deployment mode. |
| The connector does not appear in the Directory | System Applications → OmniConnector → View status | The pod is not running, or the connector's settings are wrong. Fix the connection settings and check the pod logs. |
| The connector shows OFFLINE | Directory | It stopped polling. Check the pod, the JDBC URL and the credentials. |
| The connector is DENIED | Directory | An operator denied it, so its requests for work are rejected. Select Allow. |
| Symptom | Where to look | Cause and fix |
| Unity Catalog scan returns no objects | Connector Runs | The credential lacks USE CATALOG, USE SCHEMA or BROWSE. See Prerequisites. |
| A Snowflake database shows no schemas, or the scan is partial | Resources | The role lacks USAGE on the schemas, or REFERENCES on tables and views. |
| Databricks SQL grants land on the wrong objects | Connector settings | Session catalog is not what you expect. Compare it with the warehouse's default catalog. |
| Redshift scan or grants are incomplete | Connector Runs | The login is not a superuser and lacks SYSLOG ACCESS UNRESTRICTED. |
| Lake Formation shows few or no databases or tables | Resources | The role lacks Lake Formation DESCRIBE, or is not a read-only data lake administrator. |
| A run is Partial | Connector Runs | Something could not be read. Open the run to see which objects were skipped and why, then grant the missing permission. |
Resources and scope
| Symptom | Where to look | Cause and fix |
| Nothing is scanned or governed | Directory → Scope | Include is empty, which is deny-all. Enter * or a pattern. See Managed resources. |
| Scheduled scans never run | Connector settings | Both schedules are off by default. Set a cron, or use Rescan and Re-sweep. See Scan schedule. |
| A resource is missing from Listing | Resources | It is Scoped-out or Dropped. Change the State filter, then check the scope and resource types. |
| A volume, function or other resource type is not listed | Scope → Resource types governed | Only the default types are governed. Add the type. |
Grants and drift
| Symptom | Where to look | Cause and fix |
| A grant was not applied | Workflows | See Troubleshooting a grant that did not apply. |
A workflow stays PENDING | Workflows | It waits on a dependency that failed or is waiting. Open the workflow and its dependencies. |
A workflow is FAILED | Workflows | A permanent error. Read the error. A permission error means the credential cannot grant or revoke on the object. |
Drift shows UNCONFIRMED | Grant Drift | The next permission sweep has not run. Run Re-sweep, or set a sweep schedule. |
Drift shows EXTRA or MISSING | Grant Drift | Someone changed the platform directly. Revoke the extra grant, or re-apply the missing one. |
| Observed Grants is empty | Directory | No permission sweep has run. Run Re-sweep. |
| Policies are not being enforced after a change | Recompile Queue | A resource is dead-lettered. Fix the cause shown in Last error, then select Re-drain. |
Advanced configuration
| Feature | Symptom | Cause and fix |
| Access Audits | No audits arrive | Check that Enable access audits is on and the schedule is not -. For Redshift, the login needs SYSLOG ACCESS UNRESTRICTED, or every fetch is held. For Databricks SQL, the principal must be a workspace admin. |
| Access Audits | Audits skip pages | Hold watermark on failure is off, so an undeliverable page is dropped. Turn it on to retry. Keep Page size times Request data cap under the Audit Server's inbound limit. |
| Lake Formation Federation | Tables are not federated | Federation needs a Self-Managed plane, event ingestion with Event API auth set to JWT, federation.uc-lf.enabled on, and a Lake Formation connector on the target Glue catalog. Tables with row filters or column masks are never federated. |
| Event Ingestion | Events do not arrive | Follow the checks: event source mapping, endpoint reachability, auth mode, connector ID and CloudFormation logs. |
| Event Ingestion | The Lambda gets 401 after a key rotation | Any key rotation invalidates the Lambda token. Download the bundle again, redeploy the Lambda and restart the connector pods. See Key rotation. |
| Event Ingestion | Change Events shows ERROR or OUT_OF_SCOPE | ERROR is retried. OUT_OF_SCOPE means the resource is outside the managed scope. See Change Events. |
| Custom Properties | A property has no effect | The name is wrong. Names are case-sensitive, and a name the plugin does not read is ignored. On a Shared connector the name must start with connector.. |
| Governance changes do not reach Omni Metadata Service | OMNI Sync | Check for FAILED or DEAD rows and read Last error. After the fix, use Re-emit. |