Hi,
While testing aeolus I've found that aeolus is mainly conceived for stateless instances (https://www.aeolusproject.org/redmine/issues/1059). Will additional functionalities for stateful images be added in aeolus?. I guess support for stateful instances would require:
- Differentiate between stopped and deleted (or terminated). - Being able to re-start an stopped instance. - Being able to delete an stopped instance (to destroy the underlying vm) - Being able to modify an existing instance (for example, assigning a different hardware profile)
I think statefull instances will be mainly required in private clouds using RHEV or VMware providers, i.e.
As deltacloud support both start/stop/destroy vms, I guess this is doable, is it in current roadmap plan?
Best regards,
Alfredo
On 06/30/11 - 09:36:31AM, Alfredo Moralejo wrote:
Hi,
While testing aeolus I've found that aeolus is mainly conceived for stateless instances (https://www.aeolusproject.org/redmine/issues/1059). Will additional functionalities for stateful images be added in aeolus?. I guess support for stateful instances would require:
- Differentiate between stopped and deleted (or terminated).
- Being able to re-start an stopped instance.
- Being able to delete an stopped instance (to destroy the underlying vm)
- Being able to modify an existing instance (for example, assigning a
different hardware profile)
I think statefull instances will be mainly required in private clouds using RHEV or VMware providers, i.e.
As deltacloud support both start/stop/destroy vms, I guess this is doable, is it in current roadmap plan?
It is something we want to support, but it isn't entirely clear how to support it at the moment.
One of the key goals of Aeolus is to make applications cloud-independent. That is, today I could launch my application on my private RHEV cloud, and tomorrow I could launch it on EC2.
However, stateful images break this mobility. If you start a stateful image on EC2, then for all purposes it is stuck on EC2 forever. To move to a different cloud, you'd have to launch the instances elsewhere *and copy over any data from EC2*. This may very well be a feature that users want, but we have to make clear the limitations of this approach.
So in short, yeah, it's something we want, but we aren't quite sure how to approach it yet.
On Thu, Jun 30, 2011 at 09:35:15AM -0400, Chris Lalancette wrote:
On 06/30/11 - 09:36:31AM, Alfredo Moralejo wrote:
Hi,
While testing aeolus I've found that aeolus is mainly conceived for stateless instances (https://www.aeolusproject.org/redmine/issues/1059). Will additional functionalities for stateful images be added in aeolus?. I guess support for stateful instances would require:
- Differentiate between stopped and deleted (or terminated).
- Being able to re-start an stopped instance.
- Being able to delete an stopped instance (to destroy the underlying vm)
- Being able to modify an existing instance (for example, assigning a
different hardware profile)
I think statefull instances will be mainly required in private clouds using RHEV or VMware providers, i.e.
As deltacloud support both start/stop/destroy vms, I guess this is doable, is it in current roadmap plan?
It is something we want to support, but it isn't entirely clear how to support it at the moment.
One of the key goals of Aeolus is to make applications cloud-independent. That is, today I could launch my application on my private RHEV cloud, and tomorrow I could launch it on EC2.
However, stateful images break this mobility. If you start a stateful image on EC2, then for all purposes it is stuck on EC2 forever. To move to a different cloud, you'd have to launch the instances elsewhere *and copy over any data from EC2*. This may very well be a feature that users want, but we have to make clear the limitations of this approach.
So in short, yeah, it's something we want, but we aren't quite sure how to approach it yet.
Forgive me for what might be a stupid question, but doesn't this imply separating out the instances from the long-term storage (a la S3/EBS)?
Rich.
On 08/02/11 - 06:35:42PM, Richard W.M. Jones wrote:
On Thu, Jun 30, 2011 at 09:35:15AM -0400, Chris Lalancette wrote:
On 06/30/11 - 09:36:31AM, Alfredo Moralejo wrote:
Hi,
While testing aeolus I've found that aeolus is mainly conceived for stateless instances (https://www.aeolusproject.org/redmine/issues/1059). Will additional functionalities for stateful images be added in aeolus?. I guess support for stateful instances would require:
- Differentiate between stopped and deleted (or terminated).
- Being able to re-start an stopped instance.
- Being able to delete an stopped instance (to destroy the underlying vm)
- Being able to modify an existing instance (for example, assigning a
different hardware profile)
I think statefull instances will be mainly required in private clouds using RHEV or VMware providers, i.e.
As deltacloud support both start/stop/destroy vms, I guess this is doable, is it in current roadmap plan?
It is something we want to support, but it isn't entirely clear how to support it at the moment.
One of the key goals of Aeolus is to make applications cloud-independent. That is, today I could launch my application on my private RHEV cloud, and tomorrow I could launch it on EC2.
However, stateful images break this mobility. If you start a stateful image on EC2, then for all purposes it is stuck on EC2 forever. To move to a different cloud, you'd have to launch the instances elsewhere *and copy over any data from EC2*. This may very well be a feature that users want, but we have to make clear the limitations of this approach.
So in short, yeah, it's something we want, but we aren't quite sure how to approach it yet.
Forgive me for what might be a stupid question, but doesn't this imply separating out the instances from the long-term storage (a la S3/EBS)?
Yep, exactly. That was the original thought; that the operating system would be totally stateless, and that the data that you cared about would be served by a stateful storage device (EBS, CloudFS, etc).
In the meantime, however, the world has gone and changed a bit. In particular, EBS backed (stateful) operating disks have become more popular, to the point where the "official" RedHat RHEL builds are EBS-only. One way we could treat these is as "stateless" EBS images, that we throw away every time. But I'm not sure if that is what we want to do. Hence my statement "we aren't quite sure how to approach it yet".
aeolus-devel@lists.fedorahosted.org