Batch Statistical API Issues

I am getting partially failed results in some batches I have submitted. These acted weird as well and were stuck on 99% completed for a long time. It reminded me of the last time I had issues with the batch statistical api (which resolved itself after days of waiting with no changes on my end) Empty results from batch statistical api - #5 by william.ray

This time it was stuck on 99% for a long time and most had results come back but checking the sqllite execution file a good amount simply show fatal processing error with nothing else describing why.

The two batches are:
*-b828cc45d82e
*-0635cf910bd9 (this one is still stuck on 99%)

I have submitted these batches with nothing different besides the request dates every week for the last 2 months without error. The last run was on 2024-10-22 which went through with no errors.

Hi @berb,
we did identify some infrastructure issues over the past several days, but these have been mitigated already.
Looking at these two specific batch requests, i see they are both completed now.
I am not sure I understand your comment correctly - are you saying that you did actually get results, but these are errornous?

Yup, the batch did eventually complete to 100% but a subset of the results were in error. The responses for that subset showed null data with a status of “FAILED” in the JSON files. The execution database table shows a status of FATAL with an error of “Processing error” for those entries. This was surprising as I have been using the same geodatabase for many batches in the past and have gotten results for the those POIs. I’m going to kick it off again and see if i’m still seeing the same thing, thanks!

That sounds strange.
If the error repeats, I would kindly ask you to provide some more details, so that we can debug on our side. What would help is:

  • stat batch task ID (similar as provided above)
  • example of the corrupted result (including the POI reference)
    You can send these to support@sentinel-hub.com

Thanks.

I sent out a batch identical to a successful one the other week and received some internal failures so that indicates something has changed. I’ll compile the data as concise as possible and send it on over, thanks!