Milleotto queue corrupted
Queue-system clients on many Milleotto nodes have been corrupted.
The problems with the queue system last week have resulted in the corruption of clients on many nodes. This affects parallel programs, even if they are run on a single processor (but still started through mpiexec), hanging the jobs or continously restarting them. The solution is to restart the clients, but it can only be done on empty nodes, because running jobs will be killed or at best restarted. Nodes that have not yet been fixed have therefore been taken out of the queue, pending the finishing of their running jobs.
This means that it is now safe to submit new jobs, but the number of available processors will be temporily reduced.
We are sorry for the inconvenience.
