Strange request caching issues across mulitple projects

Hey Dato Team, we currently have two projects that run on DatoCMS. One of them hasn’t been under active development for a couple months now, the other is very much in active development. Within the last 5 days, both have seemingly out of nowhere had critical issues with the data being returned from Dato. And in both cases the quick fix was to make a trivial change to the graphql request (re-order properties, overfetch) to get the data back to what we needed.

We discovered this after a bit attempting to debug it. The data was always correct in the Dato playground, and also when requesting it through a REST client (the trick there was probably that we weren’t using the exact same query as the project).

The sites call your api through Graphql-request which shouldn’t be caching anything, and they exist inside Next.js projects hosted by Vercel. I am inclined to believe Next and Vercel aren’t the issues though, because we deleted the Next.js cache locally and ran the site locally and had these issues, which were again immediately fixed with a minor alteration to the query – and I believe reoccurred when we tested undoing those minor alterations after.

Did anything strange happen on your end in the last several days? If not, let me know if you guys have any ideas if its something wrong on our end.


1 Like

Hello @tanner.hoisington and welcome to the community!

We didn’t make any changes to caching or response invalidation recently on our APIs
Does your project uses a middle API to cache the results from our graphQL endpoint?
If not, could you send us over at the requests (body and headers if possible) that had those problems, as well as the project link so we can take a look at our end and see if we can identify any possible issues?

Thank you!

Thanks Marcello! I have written in Slack once or twice but wanted to use the official channel here :stuck_out_tongue:

We aren’t using a middle API. Yeah I will reach out via email and try to provide what I can. It actually happened again today. Thanks for the response.