Dear Leo, If many jobs (or MPI processes) are accessing an NFS-mounted disk simultaneously, then NFS itself can become extremely bad. HTH, Sjors > Dear all, > > When extracting particles on our nfs cluster, recently we get extremely > long extraction times - similar jobs run for 30 hours instead of 30 > minutes before, and tend to crash towards the end. This does not seem to > be related to computational or I/O load on the cluster. Same thing happens > whether jobs are run interactively or in the queue, with one or many MPI > processes. > I noticed that load on CPUs running extraction seem to increase from low > numbers initially to 100% towards the middle/end of the run. There is no > memory swapping going on the nodes. > Any suggestions on how to track the bottleneck and remedy the situation > are highly appreciated. > Best, > Leo > -- Sjors Scheres MRC Laboratory of Molecular Biology Francis Crick Avenue, Cambridge Biomedical Campus Cambridge CB2 0QH, U.K. tel: +44 (0)1223 267061 http://www2.mrc-lmb.cam.ac.uk/groups/scheres