Replies: 1 comment 2 replies
-
|
Hi @mpm-meto, could you please provide the Also, it looks like the error from the Python script is not being output to the log. A colleague suggested to run the script outside of MET to see what error is being reported. To do this, you will need to add the directory in the MET install that contains the Python Embedding helper script that is likely being used in the script: |
Beta Was this translation helpful? Give feedback.
-
|
Hi @georgemccabe. The script does run from the command line with the latest python environment. Sorry I should've said! I will upload sample data and script to ftp. [but I can't! incoming has disappeared? Please advise new location/process.] |
Beta Was this translation helpful? Give feedback.
-
|
Just to expand on this further. Other wrappers, which do not have any python embedding, are running fine with MET12.0.0 and METplus6.0.0. The SEGFAULT seems to be directly related to interaction between python and PointStat. It could still be an environment issue but we're not sure how to test/check this. For example, does the order in which you load modules matter? |
Beta Was this translation helpful? Give feedback.
-
I have some python embedding code which you helped me develop that was last used in late 2022. It relied on a bit of python code written by Howard called met_point_obs.py back in Nov 2021. I don't know if that was part of a/the release at the time. Trying to rerun the same job using the METplus 6.0.0, Python 3.12 (our latest software stack) and MET12.0.0 causes it to fail. I don't know precisely where. The only clue is that it seems to be about to go into the python embedding. Whether there is something unsupported in the python code itself or some change between how the C++ interacts with the python, I couldn't say. We've tried the MET12 Betas as well but the same result. I really need to get it to work with the latest versions because I can't get the old MET versions to run because of the dependency on old python which is no longer available/supported.
DEBUG 3: Initializing MET compile time python instance: /net/project/ukmo/scitools/opt_scitools/conda/deployments/default-current/bin/python3.12
DEBUG 3: Running MET compile time python instance (/net/project/ukmo/scitools/opt_scitools/conda/deployments/default-current/bin/python3.12) to run user's python script (/home/h02/frmm/METplus/PointStat_ARM/read_doe_arm_obs_v5).
FATAL ERROR (SEGFAULT): Process 82607 got signal 11 @ local time = 2025-01-07 16:18:35Z
FATAL ERROR (SEGFAULT): Look for a core file in /net/home/h02/frmm/METplus/PointStat_ARM
FATAL ERROR (SEGFAULT): Process command line: /data/users/cfver/METBuild/MET12.0.0Beta6/bin/point_stat -v 6 /data/users/frmm/global/oper/nc/other/prods_op_gl-mn_20190616_00_024.nc PYTHON_NUMPY= /home/h02/frmm/METplus/PointStat_ARM/read_doe_arm_obs_v5.py /home/h02/frmm/METplus/PointStat_ARM/sgp30ebbrE11.b1.20190617.000000.nc /home/h02/frmm/METplus/PointStat_ARM/sgp30ebbrE12.b1.20190617.000000.nc /home/h02/frmm/METplus/PointStat_ARM/sgp30ebbrE13.b1.20190617.000000.nc /home/h02/frmm/METplus/PointStat_ARM/sgp30ebbrE15.b1.20190617.000000.nc /home/h02/frmm/METplus/PointStat_ARM/sgp30ebbrE32.b1.20190617.000000.nc /data/users/cfver/METBuild/METplus/parm/met_config/PointStatConfig_wrapped -outdir /data/users/frmm/MET/output/met_tool_wrapper/point_stat_py_embed_obs/
01/07 16:18:36.121 metplus.3dc37a20 INFO: Finished running /data/users/cfver/METBuild/MET12.0.0Beta6/bin/point_stat - took 0:00:03.042032
The job certainly ran back in 2022. I have the output to prove it! It would be nice to be able to replicate this before I move on to changing/adapting this for another purpose.
Beta Was this translation helpful? Give feedback.
All reactions