"MissingParameter: The request must contain the parameter SpotInstanceRequestId"

[CampGareth]

time="2018-11-08T13:56:13Z" level=error msg="Error #01: MissingParameter: The request must contain the parameter SpotInstanceRequestId\n\tstatus code: 400, request id: 5c622120-2804-4162-a3f1-891ea80d16ed\n" fields.time="2018-11-08T13:56:13Z" ip=82.13.116.126 latency=58.514545ms meta="map[Environment:production Version:b8f02dc Application:api]" method=POST path=/deployments/80ca53df-154c-4e41-95c2-cb7b80d08e56/events status=500 token=QgWmxEYVYASWoyuYnlNMVPZDlSNNnhBc type=api user-agent=Go-http-client/1.1

This occurred when Rosie was attempting to stop a deployment that was still starting up.

[CampGareth]

Digging into this, I think this only occurs when a user stops a deployment that hasn't had an underlying spot instance fire up yet.

This happens because whenever a user runs reco deploy stop <id> it posts a terminating event to platform. This gets stored in our DB and causes future things like starting an instance for that deployment to fail. After we've stored the event we also attempt to stop the associated instance under the false assumption that this event comes in after the associated instance has started up. If you run reco deploy stop very early then your deployment doesn't have an associated instance yet, not even an ID, so the stop request we send to AWS has no ID in it. This causes the above error message and ultimately makes a user see "unknown error occurred"

[pwaller]

Are you able to identify the line of code generating the error and link it here?

I assume the solution here is to simply avoid generating the remote call to AWS to stop the instance if we don't have a valid SpotInstanceRequestId to work with.

It seems that we should take care to understand the termination path to avoid the case that we try to kill a thing which is just coming into existence and miss it, thus leaving it running.

[CampGareth]

This is where we attempt to terminate the deployment on an incoming "terminating" event:

platform/handlers/api/deployment.go

Lines 297 to 299 in b8f02dc

	if event.Status == "TERMINATING" {
	err = d.DeployService.StopDeployment(c, dep)
	}

The actual error I think is coming from here:

platform/service/deployment/deployment.go

Lines 224 to 227 in b8f02dc

	_, err := ec2Session.CancelSpotInstanceRequestsWithContext(ctx, input)
	if err != nil {
	return err
	}