Recently the application running on one of our clusters had problems when some of the servers had problems accessing the NFS storage. The analysis showed there to be a mismatch between the QFE interfaces and the network protocol. The QFEs were operating at half duplex while the network was expecting full. The cluster servers had been rebooted earlier to facilitate a reconfig on the RAID level. I am trying to find out what was left out when the network cards(QFEs) were setup originally that would not show up for two months (these had been rebooted during the turnover process).