On Thu, 2009-12-10 at 11:54 +0000, Daniel P. Berrange wrote:
I've looked at the code in all the patches, and it all looks good to me. As others have said the main issue at hand is to get a better definition of the set of statistics we're going to have in the public struct.
Even though you have included a size_t parameter in the public API, this does not get usa free ride for the future, because the set of fields is also part of the RPC wire protocol ABI, and that is not so simple to deal with.
This is a going to be quite a popular/useful API for all of our hypervisor drivers. It definitely something we can easily implement for VMWare ESX, OpenVZ, and LXC drivers too. So we should try to get a struct definition that covers the memory stats required by all
Thanks for your review. I am working on an alternative way of marshaling the statistics. I will work on providing more concise definitions of each current statistic as well. -- Thanks, Adam