From: jjv5@psu.edu Newsgroups: comp.parallel.mpi Subject: Re: MPI as root on SGI PowerChallenge Date: Fri, 21 Aug 1998 00:31:00 -0500 Organization: Penn State University, Center for Academic Computing Message-Id: References: <35DCE5F6.B563DD17@esv1.auckland.ac.nz> This may be an obvious thing to point out, but are you sure root can rsh back to the machine? MPI requires rhost ability. Easy enough for a user, but normally not present for root. The few things I have run across that cause the "MPI cannot run " error are: 1) No rhost file (no rsh ability) 2) Execute permissions on the executable 3) setenv MPI_DIR $PWD and also some problems with arrayd on clusters of SGI's but your machine shouldn't have this problem. Jim In article <35DCE5F6.B563DD17@esv1.auckland.ac.nz>, Michael Somervell wrote: > I'm a masters student whose time to finish his thesis is rapidly drawing > to a close and I really need some help. To time my algorithms > accurately, I need to have exclusive access to multiple processors, and > the only way to do this is to have root access. All seems fine and good > until I actually try and run my programs using mpirun ("mpirun -np 8 > synchronous") when logged in as root (and no, I am not the > administrator, but he has kindly allowed me to run my algorithms as > root.) The error "MPI: could not run executable" appears every time, so > I look up in the mpirun man page and it says that this could basically > be anything and everything going wrong (note that I do NOT get any > errors logged in as myself and have been happily using it for about a > year.) So we (the admin and I have spent almost an entire day just > trying to figure this one out) mess round with some things, and it just > seems that you plain old cant run mpi when you're root (I tried "su" > when I had logged in as myself so all the environment variables should > still be the same.) Can anyone offer any help at all? > > (Its running on a 16 SMP R10000 processor SGI powerchallenge running > IRIX 6.2 btw.) > > Thanks