Nope, we're observing high latencies as well, as of about 21:00 UTC. Just adjusted some things and are seeing it come back down towards normal levels now.
Ah, I didn't know that. sqlx is not well supported today, I'm afraid.
It sends explicit Close messages that chew up an extra round trip instead of just dropping the connection and letting the origin DB garbage collect the prepared statements it generated, and that's not something we support currently. The bookkeeping for that is surprisingly complicated for a tx-mode pooler, and there would be a big performance penalty for doing it naively. I've poked at it a couple times.
Hyperdrive, Cloudflare's global connection pooler, relies on some key innovations to make your database connections work. We go deep into the details of what that looks like, in celebration of announcing its availability for Free Plan customers.
Funnily enough I didn't join Cloudflare until after it had a product name, so I kept wondering what this competing product was I kept seeing references to in code/docs. Eventually someone noticed and clued me in....
Super excited for these two to land! We've been putting a ton of work to make Hyperdrive your go-to choice, looking forward to getting all the feedback!
We tested in GCP and do support it. We also explicitly do support mysql_native_passwordmysql_native_password. I'm surprised, that's basically the happy path for us.
@thomasgauvin first bit of beta feedback here. I'll defer to you on the best way to capture details here.
I'm super excited for mysql hyperdrive. I am wondering, there was a question a year ago about read replicas. It sounds like that's not necessarily supported right now?
That's one option. It's probaly more performant to have hyperdrive for both, imo. Just point the reading configs at your replicas. Our internal teams do it that way for the moment.
For now, we have multiple hyperdrives (one for each replica), and use the cf geolocation headers to pick the right DB. But would be very excited to get rid of that custom logic and let Cloudflare route to the actual closest DB
We got a bunch of alerts last night as well, and couldn't find any issues wrong with our Postgres instances so assumed it was Hyperdrive. Any chance these can be added to Cloudflare status in the future so it's a little easier to tell if its our fault or not?
@AJR hey. Needed a little help with scaling hyperdrive.
We currently use Postgres with neon and we are currently hitting connection limits (1M requests per minute). We’ve plugged in hyperdrive and got it working but will occasionally hit connection limits.
I tried to connect TiDB Serverless with Hyperdrive. But I got an error:
Failed to connect to the provided database: Hyperdrive does not currently support: Hyperdrive does not currently support MySQL AuthSwitchRequest messages