A request was made to me to send a synopsis of a problem we're having at Nebraska be sent to this list.
We're having a strange issue at Nebraska involving our xrootd redirector + opportunistic clusters.
Here's the problem:
Worker nodes on our tusker cluster (with names like cxxxx.tusker.hcc.unl.edu) will make a redirection request to our xrootd.unl.edu redirector. These requests fail even though the requested file may exist in our Nebraska CMS tier-2 storage. If a direct
xrdcopy request from a worker to one of our xrootd doors the transfer works fine.
Broken attempt might look like:
Watching verbose copy attempts it appears that the redirection request also reaches out to fnal's xrootd redirector as well. These nodes are, naturally, behind a NAT but the doors and redirector can resolve the nodes directly.
I'm stumped. This looks like a redirection problem triggered by our particular network topology. I'm all certain that things used to work.
Carl Lundstedt
UNL