> Yes, this makes sense. I see that there are a few members of > glusterfs_ctx_t(ctx) > that needs to be handled with care. ctx assumes that there is only one volume > being managed in a glusterfs* process. It has only one volfile-server > connection, only one client_t table for connections to a volume etc. These > are some things one needs to take care while making ctx, the container > object for shared resources. Right. If we want to pursue this approach, these are the kinds of things I would have expected would need to be addressed. The question is: does it seem easier to move the *non-shared* items like these into struct glfs, or to move the *shared* items into one or more new resource-pool types? I suspect it's the first, but I haven't studied the details enough to be sure. _______________________________________________ Gluster-devel mailing list Gluster-devel@xxxxxxxxxxx http://www.gluster.org/mailman/listinfo/gluster-devel
