-
Notifications
You must be signed in to change notification settings - Fork 292
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
save.dataset - problem with MTG #2914
Comments
Thanks for the feedback! I never saw such an error unfortunately... Other reason I could think of is a broken environment... |
On Windows is 16 GB RAM. On Linux is 32 GB RAM. On Linux, I process operationally MSG data without any problems (presented on https://sat365.pl). Best regards, |
Even if this doesn't seem to be the case of memory running out, I suggest to try these two things to test are
This way both the resampler and writer are switched, so might help narrow down the source of the error. Both are more efficient than the ones now used. |
I have added: and changed code: composite = 'natural_color' Result is the same: |
The whole code is: from satpy.scene import Scene #with dask.config.set("array.chunk-size": "32MiB") warnings.filterwarnings('ignore') dir = max(pathlib.Path(srcdir).glob('*/'), key=os.path.getmtime) dirstr = os.path.abspath (dir) nazwa = os.path.basename (dir) Dataa = nazwa [74:82] fnames=[dirstr + '/*.nc'] #Natural Color plikout = 'NatCol_' + Dataa + '_' + Czas + '_Eu.jpg' composite = 'natural_color' |
Sorry, two lines corrected: |
Just as a sanity check, what's inside |
In dirstr are files dowloaded from EUMETSAT datastore by command: ls -l: |
Presented images were produced on Windows-10 by the same script ! |
Regarding dirstr, I was to fast with answer. Application eumdac produce the subfolder: Inside this subfolder are FCI chunks 29-41 listed in my earlier message. My script finds the latest subfolder in /home/piotr/EMCinput/MTG and this path is in dirstr. |
So I don't understand the error message copied a bit higher up, but at that point it sounds really strange. I think it's worth the try recreating the environment from scratch. |
When, I have unblocked warnings, I have got such warning: |
Ok, make sure you have the latest numpy, and that the right numpy is user if you have multiple installations of it (eg conda environment vs linux installation) |
I was playing with script and data and found interesting result. I have deleted all chunks without one - chunk number 38: The result was correct (originally Tif, converted to jpg): |
Execution with all warnings:
Edited by @pnuu |
Edited by @pnuu |
I don't know, why text is crossout. |
Latest contetn of the script:
Edited by @pnuu |
Enclose your code in a code block so the autoformatting doesn't mess it up. That is, add three ` characters before and after the code so it ends up looking like this:
|
Thanks for help.
|
I modified the last three entries to format the code/output properly. The only thing I see is what @mraspaud mentioned about Numpy. It could be the version is propably different to the one which the gradient search extension was compiled with and raises the warning. This shouldn't affect the code exectution though. Could one of the other segments be corrupted causing a problem in the NetCDF library? |
And execution:
|
|
Sorry, last message was from the second environment where I tried Python 10.
and different version ov numpy. The processing results are the same. |
On Windows, where everythig is runnig perfectly, I have numpy 1.26.4. |
Can you try changing the numpy version to 1.x to see if it helps? |
Installed numpy 1.26.4.
No result. Last message in Polish is again "Memory protection violated (memory dump). |
ok, so I'm still worried about the numpy binary compatibility. Would it be possible for you to install pyresample from source and rebuild the cython extension to see if it helps? |
Not that the numpy warning shouldn't be dealt with in the long run, but we used to get that all the time. I'm less concerned about that. I'm wondering if this memory protection violation is actually just coming from running out of memory on your machine. I haven't read this entire thread so sorry if this has already been discussed. |
@djhoese to summarize, the linux with 32GB memory doesn't work, while the windows with 16GB works. |
@pstruzik You showed it was working with only one segment. Have you tried running all segments one by one to see if one of them was corrupted? If not, how many segments to you need to run into the error? |
Great that you got it to work! |
First, I installed netCDF4 1.6.3 but was conflict with numpy. After installation of numpy 1.26.4 everything was OK. I am testing now different composition and areas. No problems until now ! |
Do you know what versions of numpy and netcdf4 you had when it crashed? |
netCDF4 1.7.1 |
Thanks, we will try this locally. |
Thanks for constructive discussion. |
Someone in the community tried with the netcdf 1.7.1 and numpy 2.1.1 and didn't have any problems... So maybe it's something else. Anyway, as it seems to work for you now, I'm closing this issue for now. |
Describe the bug
When I process MTG FCI data,, I cannot save composite to file. I,m getting message "Memory protection violation (memory dump)".
To Reproduce
Expected behavior
File MTG.png saved to actual folder
Actual results
message "Memory protection violation (memory dump)"
Screenshots
If applicable, add screenshots to help explain your problem.
Environment Info:
from satpy.utils import check_satpy; check_satpy()
]Additional context
The same code, work perfect on Windows 10 environment.
The text was updated successfully, but these errors were encountered: