From: Ilkka Karasalo Newsgroups: comp.parallel.mpi Subject: MPICH hangs at start-up Date: Thu, 05 Nov 1998 15:09:11 +0100 Organization: FOA Message-Id: <3641B186.9576EEC7@sto.foa.se> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Since Feb 98 I have been building up and using a network of workstations (HP's of various ages running HP-UX 9.x -- 10.2 and Pentium PC's running Linux (RedHat 5.1 with the 2.0.34 kernel). Recently (Oct 28), I installed the latest MPICH distribution on two of the Linux machines. The compilation, testing and installation (by 'make' 'make testing' and 'make install') worked apparently without problems. However, attempts to start MPI with any of these two machines included in the cluster ( by 'mpirun -p4pg ...' ) fail, i.e. the 'mpirun' command hangs quietly for a very long time (several minutes) and then aborts. Doing 'top' on any of these two machines - or on the machine at which the 'mpirun' was issued - during this quiet startup period reveals that a process running the intended program has indeed been created. Moreover, this process consumes close to 100% of the CPU capacity of the machine during all of the several-minutes long startup attempt !!? The MPICH release available ca Feb 98 works as it should (all hosts in the cluster respond after no more than a few seconds). I would very much appreciate comments on this peculiar behaviour of the latest MPICH release. Also: Is a distribution copy of the previous release still available, and where can I find it ? Sincerely, Ilkka Karasalo FOA 64 KTH, MWL Enkopingsvagen 126 Teknikringen 8 S-172 90 Stockholm S-100 44 Stockholm Phone: +46-8-7063627 Phone: +46-8-7908017 Fax: +46-8-7063869 Fax: +46-8-7906122 Email: ilkka@sto.foa.se Email: ilkkak@fkt.kth.se -- Ilkka Karasalo FOA 64 KTH, MWL Enkopingsvagen 126 Teknikringen 8 S-172 90 Stockholm S-100 44 Stockholm Phone: +46-8-7063627 Phone: +46-8-7908017 Fax: +46-8-7063869 Fax: +46-8-7906122 Email: ilkka@sto.foa.se Email: ilkkak@fkt.kth.se