# feedback-and-requests


01/04/2022, 8:45 AM
Update to the below feedback: Our frustration was homemade. Our data sporadically contained NaN values which made postgres fail silently. We propose to increase logging verbosity somehow, to reveal such issues. Original feedback: Hi there, we evaluated Airbyte for our company, developed a custom connector for an initial test and are now stuck where records are fetched from the source but only a fraction of them are ingested into postgres. We have a hard time to debug what is going on. We were hoping that Airbyte could replace our quite stable and performant Python ETL scripts for the better. But from what I have experienced so far is, that Airbyte will need to exit Alpha state, before we can use it in production. My question: There are so many users here... is Airbyte running stable for those of you with larger setups? I have a feeling, that it will be an awesome tool soon, but not at the moment yet.

Deepak Shah

01/04/2022, 2:26 PM
Handling NaN and converting these to nulls would be a reasonable feature request also, most databases/sources probably wont have a comparable NaN type, and I certainly dont think youll be the only person to ever run into this. I wouldnt say this issue is completely homemade!

Jeff Crooks

01/04/2022, 3:12 PM
Hi @Jens, thanks for the feedback. Did you solve your NaN handling issue? Could you please share your source catalog and the sync logs? Does the error occur in the raw data insertion step or at the normalization step?