ImageNotFound until euca-describe-images

[Joe Gordon]

Environment: Ubuntu 12.04 with Nova Essex milestone-proposed (from source), and the bleeding edge PPA for remaining OpenStack components (https://launchpad.net/~openstack-ppa/+archive/bleeding-edge).

The first time I try to run an instance I get an error which only goes away when I do a euca-describe-images.

root@compy:/home/fubar# euca-run-instances --kernel aki-00000001 -t m1.tiny ami-00000002 --key project_503
ImageNotFound: Image aki-00000001 could not be found.

root@compy:/home/fubar# euca-run-instances --kernel aki-00000001 -t m1.tiny ami-00000002 --key project_503
ImageNotFound: Image aki-00000001 could not be found.

root@compy:/home/fubar# euca-describe-images
IMAGE aki-00000001 None (natty-server-uec-amd64-vmlinuz-virtual) available public kernel instance-store
IMAGE ami-00000002 None (natty-server-uec-amd64.img) available public machine instance-store

root@compy:/home/fubar# euca-run-instances --kernel aki-00000001 -t m1.tiny ami-00000002 --key project_503
RESERVATION r-us972xu5 project_503 default
INSTANCE i-00000001 ami-00000002 server-1 server-1 pending project_503 (project_503, None) 0 m1.tiny 2012-03-21T17:47:36.000Z unknown zone aki-00000001 monitoring-disabled instance-store

[Vish Ishaya (vishvananda)]

Not sure if need to be fixed. How do you know the id of the kernel that you wish to launch until you have described instances?

image ec2 ids are assigned on demand. You are cheating and assuming that the kernel will be called aki-00000001 because it was the first image you uploaded, but IMO this is an invalide assumption. We could work around this by forcing a list and retry on not-found, but IMO this case will never exist in a real deploy, it is only a concern in toy-cloud implementations where you happen to know the order that you ordered images.

I'm marking this won't fix for now. Feel free to reopen if you disagree.

[Brian Waldon (bcwaldon)]

Removing Glance form bug report as there isn't anything to fix there.

[Joe Gordon (jogo)]

While I agree this case will never exist in a real deploy, it is an unexpected behavior that may arise while someone is evaluating nova.

While creating a work around in nova sounds a little excessive, perhaps returning a warning to the user is enough.

[Joe Gordon (jogo)]

So my vote is to re-open

[Thierry Carrez (ttx)]

Reopening as per Joe's request. Not sure where you would put the warning though ?

[Thierry Carrez (ttx)]

@Joe: suggestion on where we could put the warning ?

[Joe Gordon (jogo)]

@Theirry, perhaps in the EC2 code where the "ImageNotFound" error is thrown? It could say something about how 'image ec2 ids are assigned on demand'

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/11855

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/11855
Committed: http://github.com/openstack/nova/commit/8527f3b77773d047272624af8abdbc9356a257f2
Submitter: Jenkins
Branch: master

commit 8527f3b77773d047272624af8abdbc9356a257f2
Author: Michael Still <email address hidden>
Date: Thu Aug 23 17:39:31 2012 +1000

    Provide a hint for missing EC2 image ids

    EC2 image ids are allocated when the images are listed for the first
    time. Remind users of this if they specify an image id which does
    not exist. Resolves bug 961451.

    Change-Id: Id7326ba32e58051dbbf25c3c940fd0c9cc611ff3