Ems post grib conversion

Questions and other topics related to UEMS 15.
Post Reply
sweensta
Posts: 12
Joined: Thu Jul 02, 2015 10:17 am

Ems post grib conversion

Post by sweensta » Tue Jan 05, 2016 4:43 pm

Hi there,

After installing UEMS and transferring an old EMS domain over, I have successfully run it on our cluster. However, ems post falls over when converting the netcdf files to grib. In the screen output the error reads

Code: Select all

  

          *  Using 5 processors to run the EMS UPP - Warp speed 5!

             1 of 13 : Writing fields from 2016-01-05_06:00:00 to GRIB -  Failed (emsupp code 0)

              !  Error converting netCDF to GRIB format - The EMS has failed to exceed your expectations again!

                Log file information:

                --------------------------------------------------------------
                                                                 : 10000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000
                00000 00000 00000 00000
                                                                 : 10000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000
                00000 00000 00000 00000
                                                                 : 10000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000
                00000 00000 00000 00000
                                                                 : 10000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000
                00000 00000 00000 00000
                                                                 : 10000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000 00000
                00000 00000 00000 00000

                   ****************************************** That's the Entire List ***********************************************


                  Success - Your EMS guardian angel has left you a GRIB file:     EMSPRS.Grb2F0000000

                --------------------------------------------------------------


                  There was a problem in the creation of the grib file!

                  Try manually running the command:

                     /net/cdm1/home/user/uems/util/mpich2/bin/mpiexec  -n 5  /net/cdm1/home/user/uems/bin/emsupp < emsupp.in

                  from the emsprd/grib directory to diagnose the problem.

When I run the command as asked I get the following output

Code: Select all

[mpiexec@cdm1] HYDU_sock_listen (utils/sock/sock.c:40): error parsing port range
[mpiexec@cdm1] HYDU_sock_create_and_listen_portstr (utils/sock/sock.c:646): unable to listen on port
[mpiexec@cdm1] HYD_pmci_launch_procs (pm/pmiserv/pmiserv_pmci.c:115): unable to create PMI port
*** glibc detected *** /net/cdm1/home/user/uems/util/mpich2/bin/mpiexec: free(): invalid pointer: 0x00000000004df9a0 ***
======= Backtrace: =========
[0x4f6c3a]
[0x4f98db]
[0x45d658]
[0x40c7b9]
[0x401f3f]
[0x4df20b]
[0x400489]
======= Memory map: ========
00400000-005bb000 r-xp 00000000 00:28 55980541491                        /net/cdm1/home/user/uems/util/mpich2/bin/mpiexec
007bb000-00860000 rwxp 001bb000 00:28 55980541491                        /net/cdm1/home/user/uems/util/mpich2/bin/mpiexec
00860000-0087b000 rwxp 00000000 00:00 0
02429000-0244c000 rwxp 00000000 00:00 0                                  [heap]
2b30dd5e4000-2b30dd5e5000 rwxp 00000000 00:00 0
2b30e0000000-2b30e0027000 rwxp 00000000 00:00 0
2b30e0027000-2b30e4000000 ---p 00000000 00:00 0
7ffd7a789000-7ffd7a7ac000 rwxp 00000000 00:00 0                          [stack]
7ffd7a7c2000-7ffd7a7c4000 r--p 00000000 00:00 0                          [vvar]
7ffd7a7c4000-7ffd7a7c6000 r-xp 00000000 00:00 0                          [vdso]
ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0                  [vsyscall]
Aborted


Has anyone else come across this problem?

After some investigation, I am not sure that the bug messages we are seeing are the real problem. UEMS post-proc seems to find some error and output the messages about running .../mpiexec ... but I am not sure that line is ok or whether the messages we are getting after running that line are the same problem that UEMS is seeing.

Any thoughts much appreciated.

smartie
Posts: 97
Joined: Sat May 21, 2011 7:34 am

Re: Ems post grib conversion

Post by smartie » Sun Jan 10, 2016 9:31 am

Is the machine name set properly in the post conf file??
EDIT:
and in run_ncpus.conf see lines 287 onwards as you appear to have a MPI network interface problem.

gisk2015
Posts: 2
Joined: Tue Dec 29, 2015 10:09 am

Re: Ems post grib conversion

Post by gisk2015 » Wed Feb 03, 2016 10:23 am

Hello,

I believe your problem lies in the fact that you copied a domain from EMS. Did you replace the emsupp_cntrl.parm file with the one from UEMS?

Robert stated that the UPP has been massively reworked and in the manual for the post_processor I believe it states that the old UPP parameter file is not compatible with the new UPP.

I suggest you try to create a new test domain within UEMS to ascertain where the problem lies.

Regards.

sweensta
Posts: 12
Joined: Thu Jul 02, 2015 10:17 am

Re: Ems post grib conversion

Post by sweensta » Wed Feb 03, 2016 10:51 am

Hello,

Thankyou smartie and gisk2015 for your replies. gisk2015 you are correct. It is funny, I was just about to reply and say that we have solved the problem but you beat me to it!

We created a fresh domain using dwiz and then manually edited the namelist etc so that the new domain configuration matched our old domain. Now the grib conversion works. We now know that this is due to changes in the UPP.

Many thanks.

Post Reply