What causes SPOT_INSTANCE_TERMINATION?

The cloud provider reclaimed spot capacity due to increased demand in the region or availability zone. The spot price exceeded the maximum bid price configured for the cluster. The availability zone ran out of the requested instance type. A large batch of spot instances was reclaimed simultaneously during a cloud provider capacity rebalancing event

How do I fix SPOT_INSTANCE_TERMINATION?

Configure the cluster with on-demand fallback so jobs continue on on-demand instances if spot is unavailable. Use an instance pool with mixed spot and on-demand instances to reduce termination impact. Configure Databricks to use multiple availability zones to spread the risk of spot reclamation. For critical jobs, use on-demand instances instead of spot to avoid interruption. Enable automatic retry on the job to resume from a checkpoint if spot termination occurs

Medium severitycluster

Power BI Refresh Error:
SPOT_INSTANCE_TERMINATION

What does this error mean?

One or more spot (preemptible) instances in the Databricks cluster were reclaimed by the cloud provider. This is an expected behavior with spot instances — the cloud provider can take them back at any time when capacity is needed elsewhere.

Common causes

1The cloud provider reclaimed spot capacity due to increased demand in the region or availability zone
2The spot price exceeded the maximum bid price configured for the cluster
3The availability zone ran out of the requested instance type
4A large batch of spot instances was reclaimed simultaneously during a cloud provider capacity rebalancing event

How to fix it

1Configure the cluster with on-demand fallback so jobs continue on on-demand instances if spot is unavailable
2Use an instance pool with mixed spot and on-demand instances to reduce termination impact
3Configure Databricks to use multiple availability zones to spread the risk of spot reclamation
4For critical jobs, use on-demand instances instead of spot to avoid interruption
5Enable automatic retry on the job to resume from a checkpoint if spot termination occurs

Frequently asked questions

Can I prevent spot instance termination?

No — spot termination is a fundamental property of spot/preemptible instances. You can mitigate the impact by using fallback configuration or checkpointing.

Do Databricks jobs automatically resume after spot termination?

Only if the job is configured with retries or uses a cluster pool. Without these, the run will fail with SPOT_INSTANCE_TERMINATION.