On Tue, Jan 17, 2017 at 15:58:09 -0800, ashish mittal wrote: > On Mon, Jan 16, 2017 at 1:10 AM, Daniel P. Berrange <berrange@xxxxxxxxxx> wrote: > > On Sat, Jan 14, 2017 at 10:31:05AM -0500, John Ferlan wrote: [...] > > The pre-existing drivers in libvirt need to support both new & old syntax, > > in order that we have compat with older QEMU versions. As a brand new driver, > > IMHO, VXHS must support the new syntax exclusively - there's no reason to > > use the old syntax. We don't need to convert all existing drivers to the > > new syntax at the same time, but the VHXS patch can start the trend. > > > > Sometime back, during the qemu patch discussion, we had agreed to have > support for both URI syntax and also the QAPI syntax within qemu. It > was also decided to limit the URI syntax to supporting only one host > server, and use the QAPI exclusively for more than one servers. > > Now that the failover code has moved from qemu to libqnio, vxhs will > accept only one server. Failover will be handled entirely within > libqnio. Given that gluster is the only network protocol that supports > the new syntax, and that too only for the case when more than one > servers are specified, would it be acceptable if vxhs goes with the > URI syntax for now? We could add support for the new syntax to vxhs > later when doing it for the other protocols! No, there's no reason to do only the old syntax now, since this storage system was not present in libvirt yet. The only reason why gluster has both is that we have to support old configurations with qemu. Since there's nothing to keep running, just do the new syntax right away without any fallback or legacy code. Peter
Attachment:
signature.asc
Description: PGP signature
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
