cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From David Nalley <da...@gnsa.us>
Subject Re: Drop OVM in 4.0?
Date Thu, 18 Oct 2012 19:48:04 GMT
On Thu, Oct 18, 2012 at 3:37 PM, Joe Brockmeier <jzb@zonker.net> wrote:
> On Thu, Oct 18, 2012 at 11:10:35AM -0700, Kevin Kluge wrote:
>> There's nothing in CloudStack to track that so we cannot be definitive.
>> But, it can't be many.  We have seen few questions about its usage
>> and integration.  And OVM's share of the server virtualization market
>> is quite low.   Given limited user impact, if this is really the only
>> problem I'd defer the fix.
>
> The thing is, even if it's a small percentage of users that have
> deployed CloudStack + OVM, we'd be disappointing the shops that have
> deployed CloudStack + OVM 100%.
>
> If we were discussing why we should phase out support for OVM, it'd be
> one thing - but what seems to be on the table right now is letting
> 4.0.0-incubating out the door with no guarantee that we'll address OVM
> support in a timely fashion after, if at all.
>
> If the project is going to phase something out, we need to say so
> clearly and loudly ahead of time so interested parties have the
> opportunity to get involved and take over the feature. If nobody does,
> that's fine - but right now I'm concerned we're going to be letting down
> the users who have adopted CloudStack with OVM who had a reasonable
> expectation that the 4.0.0-incubating release would include OVM support.


So not to beat the testing drum incessantly, but to slightly diverge
from your point: Feature decisions shouldn't be decided by atrophy or
entropy. They should be conscious decisions taken by the project. To
Will's earlier comment that OVM has apparently been broken since
3.0.0, suggests that our feature list exceeds our testing capability.
We've had this same problem with other features back in 2.2.x - they
were deemed a good idea for one release, and were not tested again,
and proceeded to be unknowingly broken until someone decided to try it
and found that it didn't work (the original EIP/ELB implementation).
The cost of a feature is larger than initial development.

In the end though, no one has stood up and volunteered to own the OVM
hypervisor (even if not for 4.0); unless that changes I assume OVM is
dead within CloudStack.

--David

Mime
View raw message