Seg fault in sampleExtraction step
I got a seg fault in the sampleExtraction step. Here is the error log in the logs/sampleExtraction directory :
/var/spool/PBS/mom_priv/jobs/3281637[3].admin01.SC: line 22: 6564 Segmentation fault (core dumped) python /work/scratch/tanguyy/conda/envs/iota2-env/lib/python3.6/site-packages/iota2/Iota2.py -param_index $PBS_ARRAY_INDEX -config /home/qt/tanguyy/eolab/tanguyy/OSO_Guadeloupe/IOTA2/IOTA2_Guadeloupe_run16.cfg -starting_step 11 -ending_step 11 -config_ressources /home/qt/tanguyy/eolab/tanguyy/OSO_Guadeloupe/IOTA2/OSO_resources_2018_recommHPC.cfg
When I launch this command separately, I also get a seg fault, in the temporal Gap Filling :
2020-01-08 13:20:51 (INFO) BandMath: Image #1 (closed) has 1 components
2020-01-08 13:20:51 (INFO) BandMath: Image #2 (closed) has 1 components
2020-01-08 13:20:51 (INFO) BandMath: Image #3 (closed) has 1 components
2020-01-08 13:20:56 (INFO) ImageTimeSeriesGapFilling: Using date file /work/OT/eolab/tanguyy/OSO_Guadeloupe/IOTA2/Results_run16/features/T20QNE/tmp/Sentinel2_T20QNE_input_dates.txt 2020-01-08 13:20:56 (INFO) ImageTimeSeriesGapFilling: Using output date file /work/OT/eolab/tanguyy/OSO_Guadeloupe/IOTA2/Results_run16/features/T20QNE/tmp/Sentinel2_T20QNE_interpolation_dates.txt 2020-01-08 13:20:56 (INFO) ImageTimeSeriesGapFilling: Using linear interpolation and 10 components per date Segmentation fault
The difference with previous (successfull) runs are :
- I'm using 2017 dates instead of 2018, and there are a lot of images that don't cover the tile entirely (I did not set a relative orbit number to fetch Sentinel2 Data) -> maybe some of the images are corrupted ? How could I check it ?
I'm using Iota2 within a conda environment : I've read the discussion in #171 (closed) and tried to reinstall Iota2 in another conda environment. I've two questions :
- What is the life cycle of Iota2 in this conda environment ?
- How can we know if our installation needs an update ?
Nethertheless, I still got the seg fault with the new installation.
I also had a look to the sample selection (the ___stats.xml file) and theses stats are not empty : it should select some samples in this tile too.
Sorry because I've different questions in that bug report, but any help will be appreciated ! Thanks !