Zarr+xarray takes more time from day to day by writing with region parameter

I have built up a monthy Zarr archive with a template of dimension 30x96x1200x800 .

And every day an Array of 1x96x1200x800 is written using to_zarr(region=).

Now I am observing an increasing time to write the same amount of data from day by day. After 2/3 of the 30 days , it takes twice as long as it took on the first day.

Does anybody know why this happens? And maybe has an explanation for that?

Thanks and best regards
Daniel

It has to read and rewrite the whole chunk every time you update because your chunksize 30x… Is bigger than the update size 1x…

2 Likes

Thanks for the answer. That really helped me to understand the issue.