Mosaic
Current master of meas_mosaic (f4b19d1) built with w_2017_25 is used. All 3 tracts completed.
Coadd
Compared to DM-10129, not every patch has coadd products generated with w_2017_25. Now there are more calexps being removed from coaddition because "because size scatter is too large" or "because median e residual too large".
COSMOS: In tract 9813, the coadd products have 75 patches in HSC-G, 74 in HSC-R, 79 in HSC-I, 79 in HSC-Y, 79 in HSC-Z, and 76 in NB0921. The empty patches are:
- HSC-G: 0,0^8,0^0,1^0,7^0,8^3,8
- HSC-I: 0,0^0,8
- HSC-R: 0,0^0,8^8,0^0,1^0,7^1,8^8,8
- HSC-Y: 0,0^0,8
- HSC-Z: 0,0^0,8
- NB0921: 0,0^0,7^0,8^1,8^8,8
Processing with stack w_2017_17 in DM-10129 did generate coadd files in 0,1 and 3,8 in the HSC-G filter. Otherwise the same.
WIDE:
In tract 8766 filter HSC-R, patches 7,8 and 0,8 are empty.
In tract 8767 filter HSC-R, these patches are empty: 8,7^0,8^1,8^2,8^8,8
MultiBand
Due to the bug of DM-11091 from DM-9998, multiBandDriver would attempt to do coadd source detection on empty patches and fail because no coadd data exist in those patches. The error messages were like this: "NoResults: No locations for get: datasetType:deepCoadd dataId ...")
For this ticket, I am continuing by only doing the multiband processing for patches where coadds of all filters exist. That is, specifying the safe patch IDs only in the multiBandDriver command line --id.
In the single frame processing, among the 320 visits (103 ccds each, as ccd=9 was excluded), there were 81 reproducible fatal failures. Their dataIds are:
Out of the 81 failures: