2005-10-25 Network Problems on Docenten
A bad driver for the network cards on Docenten nodes is causing job failures
A number of nodes, mainly running parallel jobs, have crashed during a communication stage. We have diagnosed the problem to be an unstable driver for the network card. The solution is to replace it with the a driver from the card manufacturer. However, this requires several steps and before it can be applied to the whole cluster, it has to be tested. In other words, we ask for some patience.
Furthermore, nodes have to be rebooted for the new driver to take effect. When the tests have been successful a scheduled stop will be announced.
