[mvapich-discuss] PMI_KVS_Get error in mvapich2-1.5.1p1

[Mike Heinz]

I've encountered an odd piece of code while looking into this error:

Is the call to exit(1) supposed to be there? Is it left over debugging code?

pmi_tree.c:

    switch (strlen(command)) {
    case 3:                     /* get, put */
        if (0 == strcmp(command, "get")) {
            char *kvc_val = check_kvc(key);
            hdr.msg_rank = rank;
            if (kvc_val) {
                sprintf(resp, "cmd=get_result rc=0 value=%s\n", kvc_val);
                hdr.msg_len = strlen(resp);
                if (src == MT_CHILD) {
                    write(fd, &hdr, msg_hdr_s);
                }
                writeline(fd, resp, hdr.msg_len);
            } else {
                fprintf(stderr, "mpirun_rsh: PMI key '%s' not found.", key);
                exit(1);
                /* add pending req */
                save_pending_req(rank, key, fd);
                /* send req to parent */
                send_parent(rank, msg, msg_len);
            }
        }



-----Original Message-----
From: mvapich-discuss-bounces at cse.ohio-state.edu [mailto:mvapich-discuss-bounces at cse.ohio-state.edu] On Behalf Of Mike Heinz
Sent: Friday, October 15, 2010 2:30 PM
To: mvapich-discuss at cse.ohio-state.edu
Subject: [mvapich-discuss] PMI_KVS_Get error

Hi; I'm trying to diagnose a problem on a small cluster, and I'm hoping someone can point me in the right direction.

We have a standardized set of test scripts used to verify cluster operation, and on one particular cluster mvapich2 is failing with the following message:

mpirun_rsh: PMI key 'MVAPICH2_0001' not found.[cli_0]: readline failed

Googling around produced nothing useful - can someone suggest where I should look for this problem? Other MPIs are working correctly.


_______________________________________________
mvapich-discuss mailing list
mvapich-discuss at cse.ohio-state.edu
http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss