open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear Users,

I have a questions on Fortran error that has started occur recently. I am not sure if it is a specific fortran subroutine, as it indicates, or it is just my cluster fortran error. It appears when I try to merge the result files in parallel. I have been running the cases in parallel for more then one year now and have never got this error. I have not changed anything in main source files as well. Here is the error. I hope there is a fix for this as it does not depend on any specific case. I have run different simulations and sometimes I get this error. But if I rerun the same case then I might not get this error, what makes it odd. Just my cases now could take 1 month to rerun and I cannot afford to wait for it again as there is might not be garantee that it will work. Please advise. Thanks in adance.

... reading module dictionary
... running in English with cas1
recollectioning: T3DRES
At line 60 of file /home/violeta.moloney/opentelemac/v6p3/sources/utils/gretel/gretel_read_dataset.f (unit = 10, file = 'T3DRES00007-00000')
Fortran runtime error: I/O past end of record on unformatted file
... The following command failed for the reason above



Kind Regards!

Violeta

Hello,

it looks like your simulation did not complete and that the last record was not written in full (or your file not closed properly).

In any case, you can merge your simulation results without having to re-run it.

First method using gretel:
Just type again the exact same command as if you were running your simulation and add the --merge option. If you have used the -w option previously that should work right away. If you have not used the -w option, you will have to add the name of the tmp directory: -w your_tmp_dir_with_date_and_time.

Second method using the python script:
runSELFAFIN.py merge --parallel geofile rootname outname
where geofile is your original (full) geometry file, rootname can be T2DRES for instance and outname is your final output file name.

Hope this helps,
Sébastien.

Thank you very much Sebastien for your advice. I will do try it tomorrow morning as my home access to the cluster is very slow. I hope it will help. Thanks again.

Kind Regards!

Violeta

Dear Sébastien,

Thanks again for your advice. Unfortunatelly, with my cluster there are always complications . I have tried both gretel and runSelafin, and still got errors. Please see below.

1) gretel: If I type on command line:
telemac3d.py -w ./trace cas1 --merge
then I get error as before:
... reading module dictionary
... running in English with cas1
recollectioning: T3DRES
At line 60 of file /home/violeta.moloney/opentelemac/v6p3/sources/utils/gretel/gretel_read_dataset.f (unit = 10, file = 'T3DRES00007-00000')
Fortran runtime error: I/O past end of record on unformatted file
... The following command failed for the reason above


2) python: if I type on command line:
runSELAFIN.py merge --parallel sel10perc.slf T3DRES r3grn10perwallmn1pl0.0halfyear

then the error is
File "/home/violeta.moloney/opentelemac/v6p3/scripts/python27/runSELAFIN.py", line 73, in <module>
import numpy as np
ImportError: No module named numpy

Please advise if there is anything else I can try. Thanks a mil.

Kind Regards!

Violeta

Hi Violeta

It really looks like there is a problem on one of the result file...
Could you check on the temporary directory the listing of files and their size?

As a test, maybe you could try to visualize the results of this file in BK to see if you could read the file and see the results in it. Then you could compare with another piece of the result file and check if the number of time step are the same...

This is not a solution but it could helps us to understand where the problem is.

Kind regards

Thanks Christophe, I will do so. I have just noticed that it happens to most of my result files I have run for last few weeks. That is too much for coincidence. And all cases are different. I have not made any changes in source files either, so it is very strange to start getting it now. I will try to view each part of result files and will see where it will lead to. Thanks again. I will keep you posted.

Kind Regards!

Violeta

Hi Christophe,

this is strange one. I have checked the size of each result piece files, in total 7, and they all finished in full. It is 10 days simulation in model time. However, the merged file contains results for 13h and 30min only. I was able to visualize and animate the T3DRES for 13hours and a half only, but each peace of result files has results for all 10 days. Please see attached screen print where the each file size is shown. So looks like all pieces have finished but merging is not working properly. I have managed to merge other two cases minute ago and looks like for some cases it is working. If you have any thought or ideas please let me. Thanks a mil.

Kind Regards!

Violeta

Dear Violeta

I'm not sure because I cannot see the real size of each file but I suspect a problem of file size.
For example on windows 32bits machine, you cannot have file greater than 2Gb
If each individual result file is around 1.2Gb that sound logical that you could open it without any problem but with this size, the merge result file should be around 9Gb and then this is not allowed on some computer.

Hope this helps

Hi Christophe,

Thanks for your reply. Yes, you are right, for viewing result files I am using windows 32. But merging is done on cluster in parallel, so I am not sure if size could be a problem as I was merging files of more then 35Gb before easily. Now you gave me idea:). I will contact one person who is partially looks after cluster and check just in case if size per user was reduced. I have doubts, otherwise other users would have similar problem, but who knows. I will check. But please if you have any other ideas let me know and i will check. Thanks a mil Christophe.

Kind Regards!

Vioelta