Re: NCL 6.1, Batch Jobs, and Plot Problems

Mary,

Wanted to update you on this. I believe the problem I've been
experiencing is related to the postscript file problems reported by others:

http://www.ncl.ucar.edu/Support/talk_archives/2012/3818.html

I've managed to have it run correctly once or twice using postscript
files, but then using the same script it would create empty PS files.
When I switch to PDF, the files are created properly.

In response to the request in the above thread, I performed an ldd on
ncl (I will also reply to that thread with it):

         linux-vdso.so.1 => (0x00007fff18bff000)
         libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fe5ad360000)
         libbz2.so.1 => /lib64/libbz2.so.1 (0x00007fe5ad151000)
         libXrender.so.1 => /usr/lib64/libXrender.so.1 (0x00007fe5acf47000)
         libX11.so.6 => /usr/lib64/libX11.so.6 (0x00007fe5acc09000)
         libXext.so.6 => /usr/lib64/libXext.so.6 (0x00007fe5ac9f7000)
         libm.so.6 => /lib64/libm.so.6 (0x00007fe5ac7a1000)
         libdl.so.2 => /lib64/libdl.so.2 (0x00007fe5ac59c000)
         libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00007fe5ac291000)
         libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fe5ac07a000)
         libc.so.6 => /lib64/libc.so.6 (0x00007fe5abd1b000)
         /lib64/ld-linux-x86-64.so.2 (0x00007fe5ad5aa000)
         libxcb-xlib.so.0 => /usr/lib64/libxcb-xlib.so.0
(0x00007fe5abb19000)
         libxcb.so.1 => /usr/lib64/libxcb.so.1 (0x00007fe5ab8fd000)
         libXau.so.6 => /usr/lib64/libXau.so.6 (0x00007fe5ab6f8000)

- Walter

On 21-Dec-12 10:06, Mary Haley wrote:
> Hi Walter,
>
> Thanks for providing the file. I'm just now getting around to looking at it. It's been a busy month.
>
> I'm not sure if I can reproduce the batch mode environment, but I'm hoping to look at your script and see if there's anything obvious that might cause an issue.
>
> I appreciate that you provided nice test case with a README file!
>
> --Mary
>
> On Dec 13, 2012, at 8:24 PM, Walter Kolczynski wrote:
>
>> Mary,
>>
>> I've uploaded the necessary data to the FTP server in a zipped tarball with the filename wckolczy.tar.gz. I tested this tarball on a completely separate system running NCL 6.0.0 to confirm that it works as a stand-alone.
>>
>> There is a short README with instructions to run the program, some of which are duplicated here. It should only take a minute or two to complete.
>>
>> To run in PBS batch mode (using qsub), you will first need to edit the PBS directives in ensMeanSOM_singleVariable.pl as appropriate for your system. Then run using the following command:
>>
>> ensMeanSOM_singleVariable.pltest.inp
>>
>> STDOUT will be redirected to test.inp.out
>>
>>
>> To run directly on the command line, simply use the following command:
>>
>> ensMeanSOM_singleVariable_noqueue.pltest.inp
>>
>>
>> All of the relevant plotting occurs in ensMeanSOM_singleVariable.ncl:
>> line 161: creates panels
>> line 166: calls gsn_panel to put panels on page
>> (line 170 will not be encountered for this configuration)
>> line 214: creates a new xy plot with invisible points
>> line 222: places markers on xy plot for each of four colors
>> line 225/226: draw/frame
>> line 245/247: create panels
>> line 255: calls gsn_panel to put panels on page
>> (line 259 will not be encountered for this configuration)
>>
>> Here's the uname -a for the system in question:
>> Linux garnet04 2.6.32.49-0.3-default #1 SMP 2011-12-02 11:28:04 +0100 x86_64 x86_64 x86_64 GNU/Linux
>>
>> Thank you for your assistance with this. Please do not let me impact your vacation.
>>
>> - Walter
>>
>> On 12-Dec-12 17:32, Mary Haley wrote:
>>> Can you show me the part of the code where you do the graphics? I might be able to determine the problem from this.
>>>
>>> Please note that I'm on vacation at the moment, so I may not be responding as quickly.
>>>
>>> --Mary
>>>
>>> On Dec 11, 2012, at 6:24 AM, Walter Kolczynski wrote:
>>>
>>>> Mary,
>>>>
>>>> Interactive here refers to the Unix prompt; it is still submitted to NCL as ncl scriptname.ncl.
>>>>
>>>> It works if I am using an interactive (i.e., normal) Unix prompt, but when I submit it to a PBS job queue to run unattended, I get the error. The script uses ensemble data, so it can be quite large, but I will try to pare it down to a small enough sample to test on.
>>>>
>>>> - Walter
>>>>
>>>> On 10-Dec-12 12:34, Mary Haley wrote:
>>>>> Hi Walter,
>>>>>
>>>>> Sorry you're having problems with V6.1. I don't think it has to do with your differing versions of gcc.
>>>>>
>>>>> I think this is likely due to the new changes we made to the graphics in V6.1.
>>>>>
>>>>> Can you provide a sample script (and data) so we can try running it here to reproduce the problem? You can use our ftp:
>>>>>
>>>>> http://www.ncl.ucar.edu/report_bug.shtml#HowToFTP
>>>>>
>>>>>
>>>>> Also, when you say "interactive prompt", do you mean the UNIX prompt or the NCL prompt?
>>>>>
>>>>> That is, are you typing "ncl" to get into NCL's interactive mode, and then cutting-and-pasting the script to run it, or do you run it from the UNIX command line with something like:
>>>>>
>>>>> ncl myscript.ncl
>>>>>
>>>>> ?
>>>>>
>>>>>
>>>>> On Dec 10, 2012, at 12:54 PM, Walter Kolczynski wrote:
>>>>>
>>>>>> All,
>>>>>>
>>>>>> I recently tried to update from NCL 6.0 to NCL 6.1 because I needed some
>>>>>> of the new List functionality. However, when trying to create a panel
>>>>>> plot using the new version as a batch (PBS) job using qsub, I get the
>>>>>> following error:
>>>>>>
>>>>>> cro_Text: can only retrieve current color for solid patterns
>>>>>> warning:GKS:GCA: --GKS NOT IN PROPER STATE: GKS SHALL BE IN STATE GKCL
>>>>>>
>>>>>> The error repeats for each panel on the page. The output workstation (a
>>>>>> postscript file) is created, but is empty. When I try to create a
>>>>>> non-panel plot, there is still no plot but no error is printed.
>>>>>>
>>>>>> However, this only happens in batch mode. If I run the same script
>>>>>> using the interactive prompt, it runs without error. The same script
>>>>>> also works correctly in non-interactive batch mode using NCL 6.0.
>>>>>>
>>>>>> I am using the pre-compiled, non-OPenDAP RedHat binaries for both 6.0
>>>>>> and 6.1:
>>>>>> ncl_ncarg-6.0.0.Linux_RedHat_x86_64_nodap_gcc412.tar.gz
>>>>>> ncl_ncarg-6.1.0.Linux_RedHat_x86_64_nodap_gcc444.tar.gz
>>>>>>
>>>>>> The operating system is actually SUSE Linux, but as I said the RedHat
>>>>>> binaries have worked in the past. The gcc version is 4.3.4, so the
>>>>>> problem may be related to the fact that the 6.1 binary is compiled with
>>>>>> a more recent version.
>>>>>>
>>>>>> Any suggestions?
>>>>>>
>>>>>> - Walter Kolczynski, Jr.
>>>>>> NRC Post-Doctoral Researcher
>>>>>> Naval Postgraduate School Dept. of Meteorology
>>>>>>
>>>>>> _______________________________________________
>>>>>> ncl-install mailing list
>>>>>> List instructions, subscriber options, unsubscribe:
>>>>>> http://mailman.ucar.edu/mailman/listinfo/ncl-install
Received on Wed Jan 9 05:26:52 2013