If it's Bot Fight Mode, there's only really 3 options:
1) Turn off BFM and do your own firewall rules to combat automated/unwanted traffic. 2) Use IP Access Rules to whitelist originating IP addresses (which is the only way to overrule BFM by bypassing it). 3) Go onto Enterprise + Bot Management.
Thanks so much for your help @evanderkoogh! It was an issue with the BIC. Which I missed due to the UI changes. A quick Page Rule bypassed the BIC, and all is well now!
Any news on when Workers will be able to connect to Postgres, without a Cloudflare Tunnel in front? Currently, the Postgres database connector requires Cloudflared to run in front of the database host, which is impossible for connecting to serverless platforms for obvious reasons
What Erwin says is totally correct, TCP connections are required. For now if you really need Postgres you can use a provider like Supabase or self-host PostgREST, both of which provide a REST API for Postgres.
The other challenge however is that Postgres creates a transaction context per connection. So if you open up a lot of connections, you kill your Postgres instance with OOM
I'm currently using Prisma's dataproxy, but there's incompatibility with my fullstack framework of choice Thanks for pointing out that Prisma's dataproxy is the only option, I will consider changing fullstack framework
Yeah, if you must use CockroachDB or PlanetScale then that's the only way, although I have heard both of those providers are looking into how best create an HTTP API on their end.
I'm very skeptical about Supabase's claims about their scalability, because it's pure Postgres underneath, and PG requires elbow grease to scale (it's far from the predictable scaling model that NewSQL databases such as CRDB offer I know CF uses Postgres internally, so perhaps I'm wrong
Yes, and out of the two, CRDB IMHO taking the cake because it offers referential integrity while Planetscale doesn't Planetscale has got a great developer experience, with rollbacks and branches, but my use case greatly benefits from referential integrity I doubt Prisma's emulation of referential integrity for PlanetScale is reliable, because it is emulated on the client side and isolates can be killed AFAIK I was thinking about using Deno Deploy in-between CF workers and CRDB, but Deno Deploy has some rough limits during their beta: https://deno.com/deploy/docs/pricing-and-limits
The way I understand it, the max scalability of Postgres is dependent upon how powerful the master node is. For reads you can just add more replicas but for writes it's based on the master node's power.
There's a hard limit to how powerful the master can be, albeit high, it is a fundamental brick wall Am looking for a solution that doesn't sacrifice the long term for the short term, to get more buy-in from stakeholders