I should add there’s a little bit of fuzziness - I couldn’t _immediately_ see results when changing
I should add there’s a little bit of fuzziness - I couldn’t immediately see results when changing the database CA, it took <=10m
MAWS says that changing the CA is not an immediate action
CI interpreted that in the context of it waiting for the next maintenance window (which we forced immediately), not that there’d be propagation delay
CBut that is definitely possible
Cand I’m only assuming the underlying issue is cert validation - the error message was very vague
CSure thing, I added you
KOn normal Workers it requires
node_compatDifferent thing.
KThe Pages team don't plan to add
node_compatnode-postgres- other drivers requiring
is a result of the individual drivers, not Hyperdrive, correct?node_compat
KYeah
KHyperdrive as far as your Worker is concerned is just a connection string
CI wonder what part of those libs require it + if we'll ever get a "WinterCG-native" PG driver, since I don't think(?)
connect()I don't think it actually requires them, but they're not conditionally required - esbuild sees they're there, and can't tell that they won't be used at runtime with Workers.
CI thought that's what separate workerd entrypoint solves
C^^ or could solve, again I have zero knowledge of these drivers
Knode-postgresyeah me neither, and it doesn't impact our Worker luckily, but I've seen a lot of people having the
node_compatnodejs_compat
AHow to use hyperdrive in pages functions?
JUnfortunately Pages Functions doesn’t support node_compat bundling, nor the hyperdrive binding, so I’m not aware of a simple way to get hyperdrive working with Pages, sorry
J[ERROR] A request to the Cloudflare API (/accounts/cloudflare_account_id/hyperdrive/configs) failed.
not_entitled [code: 10000]
i got this error while using wrangler hyperdrive create name --connection-string="connection_string"
not_entitled [code: 10000]
i got this error while using wrangler hyperdrive create name --connection-string="connection_string"
Janyone
SAre you on the paid workers plan?
JYes
MCan you share your account ID please? (it's safe to share publicly)
J1d1c3f97856c67d63a1d69697bcb3b0d
MYou do not have a paid Workers plan.
MHyperdrive (currently) requires a paid Workers plan per https://developers.cloudflare.com/hyperdrive/get-started/#prerequisites / https://developers.cloudflare.com/workers/platform/pricing/#workers
Jthis is the test account where we test the things before the production
Jso i think we need to login with that account through wrangler
MYep. You can use
CLOUDFLARE_ACCOUNT_IDWrangler optionally uses a wrangler.toml configuration file to customize the development and deploying setup for a Worker.
Jthx it's works
HHey team, I'm integrating Hyperdrive on an existing project and it works great.
However when I run the tests using
Any ideas what I could be doing wrong ? Here's how I init the test worker.
Which results in the following error.
However when I run the tests using
unstable_devAny ideas what I could be doing wrong ? Here's how I init the test worker.
Which results in the following error.
HAFAIK
unstable_dev
MWill hyperdrive support Microsoft Sql Server at some point?
Malso I don't really understand how or why hyperdrive would make uncached queries faster.. it doesn't make sense. Especially not " 6 - 8x faster on uncached queries"
KThe blog where I assume you got those numbers from explains it all - https://blog.cloudflare.com/hyperdrive-making-regional-databases-feel-distributed/
Mso merely connection pooling is doing this? But I think most poeple of course use connection pooling without cloudflare too
KHyperdrive is specifically for Workers, which are stateless and generally short-lived.
KYou can’t hold a connection open in a Worker, it is scoped to that request. Hyperdrive gives you regional connection pools.
Khttps://blog.cloudflare.com/workers-tcp-socket-api-connect-databases/ talks about it briefly, as the TCP connect API is what Workers (and database drivers for Workers) use to talk to databases.
KThere’s also the Cloudflare backbone, which you can use yourself generally with Argo Smart Routing, which is self-described as Waze for the internet.
It’s constantly probing around and finding the best network path from A to B, and usually results in a pretty large latency decrease for cross-region requests.
It’s constantly probing around and finding the best network path from A to B, and usually results in a pretty large latency decrease for cross-region requests.
KIn the same region, sure, ASR is more like 5-7% - but we saw improvements of 55% from US and APAC to an origin in the UK,
Add in connection pooling, and query caching, and you have Hyperdrive.
Add in connection pooling, and query caching, and you have Hyperdrive.
SProbably asked before, but couldn’t find it in the chat history:
Is there any ETA of this becoming General Available? So out of beta? Thanks.
Is there any ETA of this becoming General Available? So out of beta? Thanks.
MIt’s about ~4 weeks into beta, so won’t be GA until next year. Hold tight 