Closed akremin closed 1 week ago
from a quick digging:
exposures-daily.csv
, 81096-99 are sv2
, whereas 81100-15 are special
; I confirm that with checking the header of the fiberassign-TILEID.fits.gz
file in the $DESI_ROOT/spectro/data
folder;tiles-sv2.ecsv
https://desi.lbl.gov/trac/changeset/633/data, with the message "Remove tileid 81096-81119 from tiles-sv2; these tileid were reused for fiberassign tests and are not sv2 tiles."; so I don t know... I d have to dig more!code-wise:
I ll have a look with fresh brains tomorrow, but I see that the error is in the create_petalnz_pdf()
function.
in the current code, are only considered in that function tiles from: sv1
, sv2
, sv3
, main
.
an easy / reasonable fix could be to just remove sv2
from that list, as sv2
really was more for operational tests, no science tests, if I remember correctly.
(for instance, here, the crash is in the function get_tilecov()
, which is used for the lya-diagnosis).
Would it be possible to check if the tile is in the file and if not raise a Warning but otherwise proceed without that specific tile?
sure I could do that.
that s a bit "less disruptive", even if I m pretty sure those petalnz_pdf
diagnoses are useless for sv2
...
I ll code that up, with the code reading the $DESI_SURVEYOPS/ops
folder (ie trunk
in the daily operations; maybe some tagged version for e.g. kibo).
Resolved with Pull Request #2361
Nightqa is crashing because it assumes a tile should be in a
tiles*.ecsv
file but isn't. We should investigate whether the code or ecsv file is correct here. In any case, it may also help to have this be a non-fatal error (especially for an 80000 series tile from sv2).Log:
/global/cfs/cdirs/desi/spectro/redux/kibo/run/scripts/tsnr_nightqa/logs/nightqa-kibo-20210511.log
Script:
/global/cfs/cdirs/desi/spectro/redux/kibo/run/scripts/tsnr_nightqa/scripts/nightqa-kibo-20210511.slurm
Command:
desi_night_qa --nproc=64 --recompute -p kibo -n 20210511 -o /global/cfs/cdirs/desi/spectro/redux/kibo/nightqa/20210511