Hi Nils, It might be awhile before this change can happen. In the mean time, the work-around is (1) mount using UDP, run something to force client to contact lockd, then (2) umount and remount with TCP.
-Dai Nils Goroll wrote: > Hi Dai, > >> Nils, it worked for me: >> >> pnfs-18-21.root[503] mount -F nfs -o vers=3,proto=udp >> pnfs-18-22:/tmp/ipoib /tmp/mnt > > This is really interesting, thanks a lot for trying to reproduce the > issue. > > Though it is interesting so see that under some circumstances lockd > does initiate the portmapper call via udp, this is not a solution for > the problem I am working on because, for performance reasons, I need > to run NFS over TCP. > > Would anyone be willing to improve the network lock manager code to > make the transport protocol for the portmapper call configurable? For > me, a simple tunable would be sufficient. > > Thank you, Nils
