
Spotgun polls the EC2 metadata service for a spot instance termination notice and then drains the node before it is terminated.
This gives Kubernetes almost 2 minutes to gracefully remove any pods running on a spot instance before it is terminated by AWS.
You can read more about spot instance termination notices here
Ensure that your spot nodes are labeled appropriately, e.g:
"node-role.kubernetes.io/spot-worker": "true"
Then deploy the manifest:
kubectl -f spotgun.yaml
Content type
Image
Digest
Size
21.8 MB
Last updated
almost 7 years ago
docker pull errm/spotgun