ec2.Waiter.ImageAvailable doesn't wait till image state is "available"

ec2.Waiter.ImageAvailable doesn’t wait till image state is “available”

for image in images['Images']:
    print("building image for %s" % image['Name'])
    instance = buildinstance(SharedServicesCreds, image)
    instance[0].wait_until_running()
    pp.pprint(instance[0].launch_time)
    newimage = instance[0].create_image(Name=image['Name'])
    pp.pprint(newimage.state)
    # wait_until_exits only checks that the image exists not that it is in state available
    newimage.wait_until_exists()
    waiter = sharedservicesec2client.get_waiter('image_available')
    waiter.wait(ImageIds=[ newimage.image_id ])
    pp.pprint(newimage.state)

building image for MY-RHEL7-INSTANCE datetime.datetime(2017, 8, 18, 17, 24, 43, tzinfo=tzutc()) ‘pending’ ‘pending’

[progre55]

I’ll just leave it here… not sure if this has been added recently, or always been there, but instead of the empty .wait_until_exists(), you can pass in a filter to it:

image.wait_until_exists(Filters=[{'Name': 'state', 'Values': ['available']}])

and it will return only after your image state has changed to available.

[joguSD]

My previous statement was only half correct, my apologies. While it is true that the waiter accessed via a boto3 resource will not refresh the resource, it does clear the cached data. This means that the next attempt to access an attribute on the resource will force it to be hydrated again. Thus the difference in behavior here ultimately comes down to accessing the waiter via a boto3.Resource or directly from botocore. Adding the waiter you’re using to the resource definition may be one route to make the behavior consistent. As a workaround for now you could manually refresh your resource after the waiter has ended.

We need to add the image_available waiter to the image resource.