[openstack-dev] [nova] entrypoint slowness

Sean Dague sdague at linux.vnet.ibm.com
Wed Oct 31 16:04:38 UTC 2012


On 10/31/2012 11:28 AM, Doug Hellmann wrote:
<snip>
> I'm not sure what you mean by "fast path loading." The name of an entry
> point does not include enough information, by itself, to find the class;
> the registry has to be scanned.
>
> Typically way to cache that is to hold on to the extension manager
> instance, since it already knows about all of the plugins. I'm not sure
> if that works for the driver manager, if it is trying to load different
> drivers, since it really only caches the one it is configured to load.

Ok, sorry for newbieness on python extension points here. :)

 From an outside perspective it seems like the current course and speed 
replaces something that works, with something that makes the tests take 
multiple times longer to run, without a clear win otherwise.

I'm pretty concerned about that run time addition, and really think that 
the proposed approach can't make the tests take more than 5 - 10% 
longer, otherwise the pain is really high. It can't actually make it 
through gate at the moment because it's hitting timeouts.

So I guess the question back to the experts is, is there a path here 
that makes this not add so much time to the test runs?

	-Sean

-- 
Sean Dague
IBM Linux Technology Center
email: sdague at linux.vnet.ibm.com
alt-email: sldague at us.ibm.com




More information about the OpenStack-dev mailing list