We found that newer versions of MPICH2 doesn't use MPD as a launcher. Try to look in the executor sandbox for the error message.
I would try mesos-hydra or github.com/nqn/gasc instead :) Niklas On 23 April 2015 at 11:23, Chong Chen <[email protected]> wrote: > Hi, > > > > I installed MPICH2-1.2 and Mesos 0.20.1, when I try to run a example MPI > job with mesos, the MPD is unable to start properly. Here is the output of > my test as follow. > > Could you please help me to fix this problem? Thanks! > > > > Best Regards, > > Chong > > > > bash ./mpiexec-mesos 127.0.0.1:5050 ./hello > > Connecting to Mesos master 127.0.0.1:5050 > > MPD_PID is Carmel-5_32927 > > I0423 11:15:55.252918 70968 sched.cpp:139] Version: 0.20.1 > > I0423 11:15:55.254184 71016 sched.cpp:235] New master detected at > [email protected]:5050 > > I0423 11:15:55.254294 71016 sched.cpp:243] No credentials provided. > Attempting to register without authentication > > I0423 11:15:55.254957 71039 sched.cpp:409] Framework registered with > 20150423-110655-16777343-5050-68155-0003 > > Mesos MPI scheduler and mpd running at Carmel-5:32927 > > Registered with framework ID 20150423-110655-16777343-5050-68155-0003 > > Got 1 resource offers > > Considering resource offer 20150423-110655-16777343-5050-68155-87 from > Carmel-5 > > Accepting offer on Carmel-5 to start mpd 0 > > Replying to offer: launching mpd 0 on host Carmel-5 > > We've launched all our MPDs; waiting for them to come up > > ...waiting on MPD(s)... > > Task 0 in state 1 > > ...waiting on MPD(s)... > > Task 0 in state 3 > > A task finished unexpectedly, calling mpdexit on Carmel-5_32927 > > Got 1 resource offers > > Considering resource offer 20150423-110655-16777343-5050-68155-88 from > Carmel-5 > > Declining permanently because we have already launched enough tasks > > I0423 11:15:56.777765 71038 sched.cpp:747] Stopping framework > '20150423-110655-16777343-5050-68155-0003' > > > > >
