Prevent node pool to delete node that is running a job

[Amadeu]

Hello,

I have a job running in GKE and this job may need run for a month.
Is there a way to guarantee that the node where the pod is allocated is not deleted during the resize of the node pool? We need this job to run only one time without restarting by being allocated to other node.

Best regards

[garisingh · Accepted solution]

Assuming you are using the cluster autoscaler, if you want to ensure that the node(s) running your job(s) are now downsized, then you can add the

"cluster-autoscaler.kubernetes.io/safe-to-evict": "false"

to your Job specs.  See this link for more info.

Also, if you are using release channels and have automatic upgrades enabled, you may want to consider using maintenance exclusions as well.