If it is constantly sending messages, it probably won't ever actually hibernate, but iirc since it i

If it is constantly sending messages, it probably won't ever actually hibernate, but iirc since it is eligible to be evicted, you wouldn't be billed for the time in between

HHello, I’m Allie!If it is constantly sending messages, it probably won't ever actually hibernate,...

Avi•6/8/24, 10:25 AM

oh fascinating, so you aren’t charged even if it doesn’t actually hibernate?

Avi•6/8/24, 10:26 AM

like it’s only billing time when you CANT hibernate?

AAvi like it’s only billing time when you CANT hibernate?

Hello, I’m Allie!OP•6/8/24, 10:27 AM

That's how I understand it. If you are doing work, processing an event, and haven't yet returned, it will bill you. If you have returned, then even if you haven't been evicted yet, you won't be billed.

Hello, I’m Allie!OP•6/8/24, 10:27 AM

Might be good to double check tho

Avi•6/8/24, 10:29 AM

that’s good to keep in mind for a possible cost optimization

Avi•6/8/24, 10:30 AM

now my challenge is figuring out how to handle managing the lifecycle of a postgres connection through a websocket request… i guess the AsyncLocslStorage API is the move here

Avi•6/8/24, 10:31 AM

just hard to wrap my brain around how e.g. errors are handled, since unlike a simple http request/response flow, the websocket flow is much more complicated

Avi•6/8/24, 10:31 AM

what if it errors out? what if the connection disconnects?

Avi•6/8/24, 10:31 AM

it’s not like there’s a top level “wait for websocket to finish” promise

AAvi now my challenge is figuring out how to handle managing the lifecycle of a postg...

Hello, I’m Allie!OP•6/8/24, 10:33 AM

This is still within the DO?

HHello, I’m Allie!This is still within the DO?

Avi•6/8/24, 10:33 AM

yes

AAvi yes

Hello, I’m Allie!OP•6/8/24, 10:33 AM

Why not just store it in the in-memory state directly?

Avi•6/8/24, 10:33 AM

“it” as in the database client object?

AAvi “it” as in the database client object?

Hello, I’m Allie!OP•6/8/24, 10:34 AM

Yeah

HHello, I’m Allie!Yeah

Avi•6/8/24, 10:34 AM

apparently it must be created and destroyed with every request context

Avi•6/8/24, 10:35 AM

In serverless environments such as Vercel Edge Functions or Cloudflare Workers, WebSocket connections can't outlive a single request. That means Pool or Client objects must be connected, used and closed within a single request handler. Don't create them outside a request handler; don't create them in one handler and try to reuse them in another; and to avoid exhausting available connections, don't forget to close them

Avi•6/8/24, 10:36 AM

that’s perfectly clear for a regular http request / response pair

Avi•6/8/24, 10:36 AM

and clear as mud for a websocket flow

Hello, I’m Allie!OP•6/8/24, 10:38 AM

Oh, the I/O boundary...

Hello, I’m Allie!OP•6/8/24, 10:38 AM

Uh...

Hello, I’m Allie!OP•6/8/24, 10:38 AM

Let me check something...

HHello, I’m Allie!Uh...

Avi•6/8/24, 10:39 AM

yes that’s exactly the error i was running into. i hadn’t come across this limitation before

AAvi yes that’s exactly the error i was running into. i hadn’t come across this limit...

Hello, I’m Allie!OP•6/8/24, 10:40 AM

Do you see the issue in Durable Objects too?

Hello, I’m Allie!OP•6/8/24, 10:40 AM

Because I'm not exactly sure how it applies to DOs specifically

HHello, I’m Allie!Do you see the issue in Durable Objects too?

Avi•6/8/24, 10:44 AM

to be clear, my app is only a DO

Avi•6/8/24, 10:44 AM

i mean there’s a minimal stub worker

AAvi to be clear, my app is only a DO

Hello, I’m Allie!OP•6/8/24, 10:50 AM

Hm...

AAvi to be clear, my app is only a DO

Hello, I’m Allie!OP•6/8/24, 11:24 AM

Here's a PoC that works for me:

nikiv.dev•6/8/24, 12:00 PM

I still don't get how to pass env variables on my workers deploys

nikiv.dev•6/8/24, 12:01 PM

I have

.dev.vars

.dev.vars

nikiv.dev•6/8/24, 12:01 PM

but thats not passed on deploy, just tested

nikiv.dev•6/8/24, 12:02 PM

I can't do this

nikiv.dev•6/8/24, 12:02 PM

because my project is open source

nikiv.dev•6/8/24, 12:02 PM

I don't want to put env vars in clear text

Nnikiv.dev I don't want to put env vars in clear text

Hello, I’m Allie!OP•6/8/24, 12:05 PM

For secrets, you do

wrangler secret put <SECRET_NAME>

wrangler secret put <SECRET_NAME>

nikiv.dev•6/8/24, 12:05 PM

can i have a file

nikiv.dev•6/8/24, 12:05 PM

so its declarative

nikiv.dev•6/8/24, 12:05 PM

like some file i git ignore

nikiv.dev•6/8/24, 12:05 PM

that is read from wrangler deploy

nikiv.dev•6/8/24, 12:11 PM

wrangler secret put <SECRET_NAME>

wrangler secret put <SECRET_NAME>

worked tho

Nnikiv.dev that is read from wrangler deploy

Hello, I’m Allie!OP•6/8/24, 12:15 PM

Nope.

Hello, I’m Allie!OP•6/8/24, 12:15 PM

Or well

Hello, I’m Allie!OP•6/8/24, 12:15 PM

You can suggest it in #workers-feature-suggestions

neoney•6/8/24, 5:10 PM

Using this - what is

script

script

neoney•6/8/24, 5:10 PM

Is it the name of the worker I'm trying to add the route to?

Nneoney Is it the name of the worker I'm trying to add the route to?

Hello, I’m Allie!OP•6/8/24, 5:21 PM

Yes

codebam•6/8/24, 7:14 PM

when I proxy content on my website using a worker I noticed the cache % is close to 0. is there a way to do this without breaking cache?

https://github.com/codebam/proxy-worker

GitHub

GitHub - codebam/proxy-worker

Contribute to codebam/proxy-worker development by creating an account on GitHub.

neoney•6/8/24, 8:27 PM

ok, thanks

HHello, I’m Allie!Here's a PoC that works for me:

Avi•6/8/24, 9:09 PM

i was doing basically the same thing, but storing the reference on the global scope. i wonder if having it be an instance of the DO changes things?

Avi•6/8/24, 9:09 PM

i was also creating the reference in the DO constructor, instead of lazily instantiating it like you are – why are you doing it that way, btw?

If it is constantly sending messages, it probably won't ever actually hibernate, but iirc since it i

Similar Threads