failed to query servers: failed to deallocate cached statement(s): conn busy

[loretoparisi]

Describe the bug
Error when calling the api https://registry.modelcontextprotocol.io/v0/servers?limit=10

"message": "failed to query servers: failed to deallocate cached statement(s): conn busy"

To Reproduce
Steps to reproduce the behavior:

1. curl https://registry.modelcontextprotocol.io/v0/servers?limit=10

or

open https://registry.modelcontextprotocol.io/v0/servers?limit=10

Expected behavior
retrieve the json with server list

Logs

Additional context
@domdomegg @toby

[ryanbaumann]

+1 same error when trying to publish a new server to the registry

[rodincave]

Same here when trying to publish new server via dns login

[cameroncooke]

Been like this for about 10 hours, this feels more like an outage than a transient availability issue.

[rdimitrov]

Cross-posting from Discord:

Thanks so much to everyone who’s been reporting these issues. That’s incredibly helpful and much appreciated! 🙏 It’s also really encouraging to see the level of interest from the community which speaks volumes about the value of the registry itself.

That said I’d like to highlight that this is still a preview release and while everyone is excited to see people exploring it there aren’t any hard guarantees yet about the stability of the registry at this stage. The FAQ has a bit more detail if you're interested: https://github.com/modelcontextprotocol/registry/blob/main/docs/reference/faq.md#whats-the-expected-reliability

[loretoparisi]

Thanks @rdimitrov, my contribution/suggestion here would be to prioritize the discovery api above, that is the entry point of the system! Thanks a lot for the great work.

[domdomegg]

Hey folks, sorry for the outage here! We've been really excited to see the high demand for the registry even as it's in preview :D

@rdimitrov I think has pushed a fix for the issue in #408, which we're deploying and monitoring now.

Apologies again for the outage, and I'll post another status update once we've confirmed the API errors have gone down.