COGs built from VRT don't validate about rio-cogeo HOT 10 CLOSED

Thanks for the issue @drwelby , any chance you can share the vrt (+tiffs) or create dummy data ?

from rio-cogeo.

Yes, I should be able to copy some data into a public bucket.

from rio-cogeo.

rio cogeo create /vsis3/rio-cogeo-vrt-sample/rio-cogeo-s3.vrt ./rcv-test.tif -p jpeg

of course this works locally, so I'll have to investigate our process container to see why it generates invalid COGs.

from rio-cogeo.

@drwelby could you check rasterio and gdal version ?

from rio-cogeo.

rasterio==1.0.26
rio-cogeo==1.1.0
pygdal==2.1.2.5

libgdal20=2.1.2
libgdal-dev=2.1.2

is what is installed in the container, but I should check to make sure nothing is getting downgraded

Locally I have:

GDAL 2.4.1
rasterio 1.0.24

from rio-cogeo.

@drwelby This is strange and it seems to me /vsis3/bucket/my.tif is not the output of rio cogeo create. I've just tested and it works fine on my side (even if it really slow).

To be honest, It might be faster to do

gdal_translate my.vrt my.tif -co COMPRESS=JPEG
rio cogeo create my.tif my_cog.tif -p jpeg

from rio-cogeo.

We upgraded to GDAL 2.4.1 in the container and now valid COGs are being created.

Thanks for the suggestion of adding a gdal_translate step - we'll benchmark it and see if it provides any gains.

from rio-cogeo.

Tried it out and writing the intermediate file turned out to be ~5% slower in our setup.

from rio-cogeo.

@drwelby are we good to close here ?

from rio-cogeo.

Closing, problem on our end.

from rio-cogeo.