I have some big Spark workloads that can be run with a flexible schedule. I would like to take advantage of the low cost of the Spot instances of the cloud providers. Can I run the Spark jobs in the Iguazio platform with the Spot instances?
I have some big Spark workloads that can be run with a flexible schedule. I would like to take advantage of the low cost of the Spot instances of the cloud providers. Can I run the Spark jobs in the Iguazio platform with the Spot instances?
Yes, you can run Spark jobs in Iguazio using spot instances. To run the job on the Spot instance, first, you will need to config a new NodeGroup with Spot instances (e.g. added-spot-ng). Then in your job, specify the node selection like the code snippet below:
fn.with_node_selection(node_selector={'app.iguazio.com/node-group': 'added-spot-ng'})
This way, the function (Spark job) will be scheduled to the Spot instance node-group.