Are you able to capture the debug or panic output from the nodes before they reboot? It is difficult to diagnose the issue without that information. On 03/27/2015 10:52 AM, Vaibhav Rekhate wrote:

I have setup a cluster on Ubuntu 12.04 machines (using the steps outlined here: https://help.ubuntu.com/community/MpichCluster)

The environment variable PATH is set. The code is compiled on the master node and executed on the master node. After execution of the code, the slave nodes crash, i.e. the slaves start rebooting, debug statements on the display indicate kernel panic. The slaves have to be restarted. This is not exactly reproducible as it does not happen every time, but sometimes.

Regards, Vaibhav Rekhate B. Tech, Computer Engineering College of Engineering, Pune ------------------------------------------------------------------------ *From:* Huiwei Lu <[email protected]> *Sent:* 27 March 2015 07:41 PM *To:* [email protected] *Subject:* Re: [mpich-discuss] Slave hosts panic after mpirun Hi Vaibhav,

Can you give us some details of how the slaves panic?

-- Huiwei Lu

On Fri, Mar 27, 2015 at 2:19 AM, Vaibhav Rekhate <[email protected] <mailto:[email protected]>> wrote:

Hello there! I have setup a small cluster (only 3 nodes - 1 master, 2 slaves). I was trying to run a sample program by using this command:

$MPI_INSTALL_DIR/bin/mpirun -n 9 --machinefile machinefile ./a.out

Contents of machine file:

host1:3 host2:3 host3:3

Please help.

Regards, Vaibhav Rekhate _______________________________________________ discuss mailing list [email protected] <mailto:[email protected]> To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss