Previously, you exposed cache availability through xattr. This was useful for loadbalancing at Hadoop. Now, I need more than that. Can we expose local cache path (or parent directory of cache block files) through xattr? I found there's a function opening a cached block file (https://github.com/jcnelson/syndicate/blob/master/UG/fs/read.cpp#L374) in the source code level. But more generally, if we can get the shortcut to the cache block file, Hadoop can directly read it instead of sending the data through local socket.
Previously, you exposed cache availability through xattr. This was useful for loadbalancing at Hadoop. Now, I need more than that. Can we expose local cache path (or parent directory of cache block files) through xattr? I found there's a function opening a cached block file (https://github.com/jcnelson/syndicate/blob/master/UG/fs/read.cpp#L374) in the source code level. But more generally, if we can get the shortcut to the cache block file, Hadoop can directly read it instead of sending the data through local socket.