Qfyilyi

Before knowing exactly how AWS spot instances work I configured a spot request with the interruption behavior set to terminate. As I understand it my running instance state will be deleted on termination. So if I don't have an image backup I will not be able to start the server again in it's last state.

Since the spot request is fulfilled and the instance running, is it possible to change the interruption behavior to stop when I am outbid? I cant seem to find the option to change the interruption behavior.

According to boto3 Spot instance creation documentation, you may also instruct the instance to stop or hibernate if you set the instance type to persistent. The default behaviour is terminate. This features is added in November 2017.

response = client.request_spot_fleet(

   SpotFleetRequestConfig={

     .....

     Type='one-time'|'persistent',

     'InstanceInterruptionBehavior': 'hibernate'|'stop'|'terminate'

    }

Use them sparingly as each behaviour has some pros and cons, e.g. you must take care of process network connection interruption if yo use hibernate. For stop, you may want to store data into another mounted EBS.

For setting the interruption behavior to stop, we need to take care of some requirements:

• For a Spot Instance request, the type must be persistent, not
  one-time. You cannot specify a launch group in the Spot Instance
  request.


• For a Spot Fleet request, the type must be maintain, not request.


• The root volume must be an EBS volume, not an instance store volume.

by following the above requirements, we can change the interruption behavior from terminate to stop.

Please refer the following url for reference:
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/spot-interruptions.html