MetricSign
EN|NLRequest Access
High severitycluster

Power BI Refresh Error:
INSTANCE_UNREACHABLE_DUE_TO_MISCONFIG

What does this error mean?

Databricks cannot reach the cluster instance after it was launched. The VM started but the Databricks control plane cannot communicate with it, and the root cause is a network or infrastructure misconfiguration.

Common causes

  • 1A security group or network security group does not allow inbound traffic from the Databricks control plane to cluster nodes
  • 2The cluster subnet routing table does not have a route back to the Databricks control plane or VPC endpoint
  • 3An asymmetric routing configuration where requests can reach the instance but responses cannot return
  • 4A misconfigured network appliance (firewall, NAT) between the cluster and the Databricks control plane
  • 5A VPC peering or Private Link misconfiguration that allows the VM to launch but blocks Databricks communication

How to fix it

  1. 1Verify that inbound security group rules allow traffic from the Databricks control plane CIDR ranges
  2. 2Check routing tables in the cluster subnet for correct routes back to the Databricks VPC endpoint or internet gateway
  3. 3Test bidirectional connectivity — the cluster VM must be able to both reach out to and receive connections from the control plane
  4. 4Check for any asymmetric routing: ensure all traffic in and out of the subnet uses the same path
  5. 5Review recent VPC/VNet or security group changes and roll back if the change correlates with this error appearing

Frequently asked questions

What is the difference between this and NETWORK_CHECK_CONTROL_PLANE_FAILURE?

NETWORK_CHECK_CONTROL_PLANE_FAILURE means the cluster cannot reach the control plane (outbound blocked). INSTANCE_UNREACHABLE_DUE_TO_MISCONFIG means the control plane cannot reach the cluster instance (inbound blocked).

Does this error affect the cluster data plane?

Potentially yes — if the control plane cannot communicate with the cluster, job scheduling and monitoring fail. The cluster may technically have compute running but be ungovernable.

Other cluster errors