DescribeInstances does not raise error on invalid instance

This is due to the way that we check multiple ids and should be an easy fix. While you are at it can you also post what the response should be if there are two ids specified and one exists and the other doesn't?

Taking, will try to reproduce and work on fix

I see the problem at _format_instances:

    def _format_instances(self, context, instance_id=None, use_v6=False,
            **search_opts):
        # TODO(termie): this method is poorly named as its name does not imply
        # that it will be making a variety of database calls
        # rather than simply formatting a bunch of instances that
        # were handed to it
        reservations = {}
        # NOTE(vish): instance_id is an optional list of ids to filter by
        if instance_id:
            instances = []
            for ec2_id in instance_id:
                internal_id = ec2utils.ec2_id_to_id(ec2_id)
                try:
                    instance = self.compute_api.get(context, internal_id)
                except exception.NotFound:
                    continue
                instances.append(instance)

If no instance is found wit the ID passed as an argument, the exception is caught but we don´t print anything about it .

root@openstack-dev:/home/dormammu/src/openstack/nova# euca-describe-instances i-aaaaaaaa
InstanceNotFound: Instance i-aaaaaaaa could not be found.

root@openstack-dev:/home/dormammu/src/openstack/nova# euca-run-instances ami-tiny -t m1.tiny
ImageNotFound: Image ami-tiny could not be found.

root@openstack-dev:/home/dormammu/src/openstack/nova# euca-describe-instances i-aaaaaaaa
InstanceNotFound: Instance i-aaaaaaaa could not be found.

root@openstack-dev:/home/dormammu/src/openstack/nova# euca-run-instances ami-tty -t m1.tiny
RESERVATION r-91huht8j admin default
INSTANCE i-00000001 ami-00000008 pending None (admin, None) 0 m1.tiny 2011-09-21T10:28:16Z unknown zone aki-00000007 ari-00000006

root@openstack-dev:/home/dormammu/src/openstack/nova# euca-describe-instances i-00000001
RESERVATION r-91huht8j admin default
INSTANCE i-00000001 ami-00000008 10.0.0.3 10.0.0.3 running None (admin, openstack-dev) 0 m1.tiny 2011-09-21T10:28:16Z nova aki-00000007 ari-00000006

root@openstack-dev:/home/dormammu/src/openstack/nova# euca-describe-instances i-aaaaaaaa
InstanceNotFound: Instance i-aaaaaaaa could not be found.

Fix proposed to branch: master
Review: https://review.openstack.org/3978

Review failed and there doesn't look like there's been another fix attempt since.

I went to write a patch for this, but I can't until I know what to do in the case that two instance ids are passed and one is valid and the other is not. What should happen then?

I ran the attached program with 2 instances running to show what would happen. The short summary:
 If there are invalid instances in the request, do not terminate anything.

Full output:

original: [u'i-2046ba5e/running', u'i-0a46ba74/running']
kill_good_first start: [u'i-2046ba5e/running', u'i-0a46ba74/running']
kill_good_first: killing [u'i-2046ba5e', u'i-11111111']
caught exception
  code=InvalidInstanceID.NotFound
  msg=The instance ID 'i-11111111' does not exist
  reason=Bad Request

kill_good_first end: [u'i-2046ba5e/running', u'i-0a46ba74/running']
kill_bad_first start: [u'i-2046ba5e/running', u'i-0a46ba74/running']
kill_bad_first: killing [u'i-22222222', u'i-0a46ba74']
caught exception
  code=InvalidInstanceID.NotFound
  msg=The instance ID 'i-22222222' does not exist
  reason=Bad Request

kill_bad_first end: [u'i-2046ba5e/running', u'i-0a46ba74/running']

Fix proposed to branch: master
Review: https://review.openstack.org/18677

This change of behaviour causes tempest to fail. I need to update the devstack euca exercise.

Fix proposed to branch: master
Review: https://review.openstack.org/18701

Reviewed: https://review.openstack.org/18701
Committed: http://github.com/openstack-dev/devstack/commit/796342c06e8ca3dcfd2e8f1ba4e0300a703b8de1
Submitter: Jenkins
Branch: master

commit 796342c06e8ca3dcfd2e8f1ba4e0300a703b8de1
Author: Michael Still <email address hidden>
Date: Fri Dec 28 11:08:20 2012 +1100

    Handle the new behaviour for invalid instances.

    The behaviour of this case changed with bug/836978. Requesting the
    status of an invalid instance will now return an error message
    including the instance id, so we need to filter that out.

    Resolves the devstack elements of bug 836978.

    Change-Id: I385eb6f04cff90e1ddc0b79d835fbbdf92e4e9ff

Reviewed: https://review.openstack.org/18677
Committed: http://github.com/openstack/nova/commit/66d0cb1ce5c9d716ad93685ab6e6e86ddbd0b293
Submitter: Jenkins
Branch: master

commit 66d0cb1ce5c9d716ad93685ab6e6e86ddbd0b293
Author: Michael Still <email address hidden>
Date: Thu Dec 27 16:15:12 2012 +1100

    Invalid EC2 ids should make the entire request fail.

    Resolves bug 836978. I suspect there are many other cases where we
    should add this check.

    Change-Id: I027e44db2e27eb1ef913ddad8560cca08388906b