@nacx @andreaturli My opinion is that we should offer a consistent cloud
abstraction with jclouds. I don't think any of the other providers require the
region to be included in the hardwareId, so ideally we shouldn't require it for
GCE.
(Is that statement true? Do any other providers include the region/zone in the
hardwareId?)
But I'm not yet sure how we'd achieve that while maintaining consistency
between listHardwareProfiles and template options!
Some general thoughts/observations below.
---
As for some machine-types (i.e. hardwareIds) not being available in some zones,
that is also true for aws-ec2 (new hardware profiles are usually introduced in
just a sub-set of regions first). We are happy to use the simple name in
jclouds, but then AWS is a bad example because their API doesn't let you list
the machine-types!
Looking at how Google themselves present this in the CLI, they use the simple
name rather than asking people to repeat the zone:
https://cloud.google.com/sdk/gcloud/reference/compute/instances/set-machine-type
There is an interesting comparison for AMIs however: in aws-ec2 the AMI id
includes the region. However, I think that is not a fair comparison because the
id will always be different in different regions (or if it is the same it is
purely coincidence).
---
As to how `listHardwareProfiles` should behave, I agree that's very tricky. We
could return `n1-standard-1` multiple times, with each instance having a
different `hardware.getLocation()`. But the hardware.getId javadoc is clear
that if we did so then the id should include the zone:
/**
*
* A means to uniquely address this resource within a provider. For
example, if the namespace of
* a node or image is region based, the id will likely include both the
region and the
* provider-supplied id encoded to avoid collisions.
*
* @return unique id within your account on the provider
*/
And if `hardware.getId()` includes the zone, then for consistency that is what
a user should supply in the template.
An alternative would be to *not* tell the user which locations support which
machine-types when they call `listHardwareProfiles`. That feels consistent with
the (ugly) aws-ec2 implementation, but sub-optimal.
---
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/jclouds/jclouds/pull/951#issuecomment-217143429
