Hi,
We just got hit by a problem with sharedfp/lockedfile component under
v2.0.1 (should be identical with v2.0.2). We had 2 instances of an MPI
program running conccurrently on the same input file and using
MPI_File_read_shared() function ...
If the shared file pointer is maintained with the lockedfile
component, a "XXX.lockedfile" is created near to the data
file. Unfortunately, this fixed name will collide with multiple tools
instances ;)
Running 2 instances of the following command line (source code
attached) on the same machine will show the problematic behaviour.
mpirun -n 1 --mca sharedfp lockedfile ./shrread -v input.dat
Confirmed with lsof(8) output :
njoly@tars [~]> lsof input.dat.lockedfile
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
shrread 5876 njoly 21w REG 0,30 8 13510798885996031
input.dat.lockedfile
shrread 5884 njoly 21w REG 0,30 8 13510798885996031
input.dat.lockedfile
Thanks in advance.
--
Nicolas Joly
Cluster & Computing Group
Biology IT Center
Institut Pasteur, Paris.
#include <sys/statvfs.h>
#include <sys/utsname.h>
#include <assert.h>
#include <err.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <mpi.h>
int main(int argc, char **argv) {
char *rbuf, *file;
int i, res, cnt, dry, inf;
size_t len;
struct statvfs stv;
struct utsname nam;
MPI_Comm comm = MPI_COMM_WORLD;
MPI_Info info = MPI_INFO_NULL;
MPI_File rfh;
MPI_Offset size, read, rsum;
MPI_Status sts;
dry = inf = 0;
while ((i = getopt(argc, argv, "nv")) != -1) {
switch (i) {
case 'n': dry++; break; /* dryrun */
case 'v': inf++; break; /* verbose */
default: return EXIT_FAILURE; }
}
if (argc - optind != 1)
errx(1, "usage: %s <file>", *argv);
file = *(argv+optind);
res = statvfs(file, &stv);
assert(res == 0);
len = 10UL * 1024 * 1024;
len -= (len % stv.f_bsize);
rbuf = malloc(len);
assert(rbuf != NULL);
res = MPI_Init(&argc, &argv);
assert(res == MPI_SUCCESS);
res = MPI_File_open(comm, file, MPI_MODE_RDONLY, info, &rfh);
assert(res == MPI_SUCCESS);
res = MPI_File_get_size(rfh, &size);
assert(res == MPI_SUCCESS);
read = 0;
while (1) {
res = MPI_File_read_shared(rfh, rbuf, (int)len, MPI_CHAR, &sts);
assert(res == MPI_SUCCESS);
res = MPI_Get_count(&sts, MPI_CHAR, &cnt);
assert(res == MPI_SUCCESS);
if (cnt == 0) break;
read += cnt;
assert(read <= size);
}
res = MPI_File_close(&rfh);
assert(res == MPI_SUCCESS);
rsum = 0;
res = MPI_Allreduce(&read, &rsum, 1, MPI_OFFSET, MPI_SUM, comm);
assert(res == MPI_SUCCESS);
res = MPI_Finalize();
assert(res == MPI_SUCCESS);
free(rbuf);
if (inf != 0) {
assert(uname(&nam) == 0);
printf("%s: %lld/%lld/%lld\n", nam.nodename, read, rsum, size); }
assert(rsum == size);
return 0; }
_______________________________________________
users mailing list
users@lists.open-mpi.org
https://rfd.newmexicoconsortium.org/mailman/listinfo/users
