GraphQL request failed For the past two hours or so we've been seeing data proxy server failing requests intermittently. We didn't have any deploy making changes in the data proxy code. Do you know if Data Proxy is having a partial outage?

Are you getting a specific error code back @Rodrigo Oliveira for the failed requests? I am thinking it could be perhaps that you are hitting limits?

@Alberto Perdomo We are not getting any specific code back, just GraphQL request failed for like 5 requests out of 30-50 per minute we run

I am not aware of any partial outages, but let me look into this.

@Alberto Perdomo Thanks, let me know if you need more information

@Alberto Perdomo does the data proxy use graphql under the hood or why would we get graphql errors in proxy.js?

The Client communicates with the Query Engine using GQL.

Thanks for the update @Alberto Perdomo!

CC @Scott Downes @Keenahn Jung

Hello, the issue has been mitigated and we are no longer seeing elevated error rates. We continue to monitor the systems to verify that the mitigation works as expected. @Rodrigo Oliveira @Pieter

We are able to confirm that our systems have been restored to normal! Thanks a lot for looking into this @Alberto Perdomo!

I am glad to hear!

I will update Prisma Client to v3.5.0 soon enough, if not today but Monday to get use of the new functionality.

I just logged into ask if anyone is using data proxy in production 😮 I'll probably not migrate my saas just yet..

@Sai Krishna It worked really well for us in the past ~2 weeks. We've got some service degradation today (5-10% of requests were failing) as we have about 30-50 rpm to the endpoints we are using data proxy with. Obviously today's partial outage affected a lot users in our platform and brought a few other services down, intermittently. Prisma team paid a lot of attention to the issues we've flagged and resolved them in a quick manner. Just giving you some insights of our experience so far for you to analyze what risks you can take and help with your decision.

Thank you for the detailed writeup @Rodrigo Oliveira 🙂 I'd still wait for some more time. Right now, I deployed my nextjs dashboard as a containerized nodejs-express app with auto-scaling. So, I don't have a connection pooling issue for a considerable amount of time.

Just wanted to provide some additional context: During the issues on Friday, it wasn’t a general issue but an issue with specific instances.

Thank you @Alberto Perdomo I will keep 👀