Why is this happening and why is no one fixing it?

K

KenKiyosakiOP•12/10/24, 2:38 PM

@Community time

K

KenKiyosakiOP•12/10/24, 2:39 PM

@cb @codefraw

Z

zegevlier•12/10/24, 2:44 PM

?pings

S

SuperHelpflare•12/10/24, 2:44 PM

Please do not ping community members for non-moderation reasons. Doing so will not solve your issue faster and will make people less likely to want to help you.

R

rivage•12/10/24, 9:35 PM

using websocket hiberation api with durable object storage. looks like

webSocketClose

webSocketClose

isn't called when the websocket isn't closed cleanly which causes the duration to build.

is it recommended to use an alarm to clean up the durable object data?

Rrivage using websocket hiberation api with durable object storage. looks like `webSocke...

1

1984 Ford Laser•12/10/24, 10:30 PM

You mustn't be using hibernation correctly as it shouldn't build up any duration if you aren't sending any messages. Are you calling

this.state.acceptWebSocket()

this.state.acceptWebSocket()

?

KKenKiyosaki Why is this happening and why is no one fixing it?

I

Isaac McFadyen•12/10/24, 10:43 PM

As has been said above, something about your framework is pollyfilling fetch -> XMLHTTPRequest. This is not a Workers issue.

I

Isaac McFadyen•12/10/24, 10:44 PM

We can't help without knowing what framework you're using, and even then we likely can't help. You'll need to talk to the framework authors.

11984 Ford Laser You mustn't be using hibernation correctly as it shouldn't build up any duration...

R

rivage•12/10/24, 10:52 PM

yes:

  async fetch(request: Request) {
    await this.ensureMessagesInitialized();
    const webSocketPair = new WebSocketPair();
    const [client, server] = Object.values(webSocketPair);
    this.ctx.acceptWebSocket(server);
    return new Response(null, {
      status: 101,
      webSocket: client,
    });
  }

  async fetch(request: Request) {
    await this.ensureMessagesInitialized();
    const webSocketPair = new WebSocketPair();
    const [client, server] = Object.values(webSocketPair);
    this.ctx.acceptWebSocket(server);
    return new Response(null, {
      status: 101,
      webSocket: client,
    });
  }

  async fetch(request: Request) {
    await this.ensureMessagesInitialized();
    const webSocketPair = new WebSocketPair();
    const [client, server] = Object.values(webSocketPair);
    this.ctx.acceptWebSocket(server);
    return new Response(null, {
      status: 101,
      webSocket: client,
    });
  }

  async fetch(request: Request) {
    await this.ensureMessagesInitialized();
    const webSocketPair = new WebSocketPair();
    const [client, server] = Object.values(webSocketPair);
    this.ctx.acceptWebSocket(server);
    return new Response(null, {
      status: 101,
      webSocket: client,
    });
  }

Rrivage yes: ``` async fetch(request: Request) { await this.ensureMessagesInitial...

1

1984 Ford Laser•12/10/24, 10:56 PM

Very odd, are you sure that usage is from WS? Any clients staying connected and sending messages repeatedly? What are users broadcasting

11984 Ford Laser Very odd, are you sure that usage is from WS? Any clients staying connected and ...

R

rivage•12/10/24, 10:59 PM

running locally, have had ~5 people test it, but i think this is due to durable object storage

R

rivage•12/10/24, 11:00 PM

the documentation says websocketclose will be called whenever the websocket closes to clean up storage

R

rivage•12/10/24, 11:00 PM

here's the implementation of the

deleteAll()

deleteAll()

:

async webSocketClose(
    ws: WebSocket,
    code: number,
    reason: string,
    wasClean: boolean,
  ) {
    // cleaning up storage if you want, or remove if you prefer not to delete
    await this.state.storage.deleteAll();
    ws.close(code, "durable object is closing websocket");
  }

async webSocketClose(
    ws: WebSocket,
    code: number,
    reason: string,
    wasClean: boolean,
  ) {
    // cleaning up storage if you want, or remove if you prefer not to delete
    await this.state.storage.deleteAll();
    ws.close(code, "durable object is closing websocket");
  }

async webSocketClose(
    ws: WebSocket,
    code: number,
    reason: string,
    wasClean: boolean,
  ) {
    // cleaning up storage if you want, or remove if you prefer not to delete
    await this.state.storage.deleteAll();
    ws.close(code, "durable object is closing websocket");
  }

async webSocketClose(
    ws: WebSocket,
    code: number,
    reason: string,
    wasClean: boolean,
  ) {
    // cleaning up storage if you want, or remove if you prefer not to delete
    await this.state.storage.deleteAll();
    ws.close(code, "durable object is closing websocket");
  }

R

rivage•12/10/24, 11:00 PM

i'm curious if i should implement an alarm to do this on an interval

1

1984 Ford Laser•12/10/24, 11:03 PM

Your issue sounds like something else. It sounds like something else is pinning the server to stay open. I think when it says clean up storage, that's for the server to handle on its own. Having storage after a server is closed won't continue to incur duration charges

1

1984 Ford Laser•12/10/24, 11:04 PM

I don't even have a distinct

webSocketClose()

webSocketClose()

in my code and I haven't had any issues with duration charges

A

akashra•12/10/24, 11:32 PM

oof. I'm still struggling with figuring out an appropriate practice/handling technique for session management and data caching to reduce hits to D1, DO or KV services - especially given the relative high cost of writes to those services to update session caches with any request

AAshley The matrix is for the core Pages vs Worker compat, so it’s talking about file-ba...

A

Abdul•12/11/24, 12:33 AM

Thanks Ashley.

Will dig deeper into these and see what i can find

M

masd4•12/11/24, 7:32 AM

EDIT: The answer is that I had to upgrade to the latest wrangler and compatability date. (((old question: hi, according to these docs tail workers are supported for workers with assets but when I try to deploy, it says:

Cannot use assets and tail consumers in the same Worker. Tail Workers are not yet supported for Workers with assets

Cannot use assets and tail consumers in the same Worker. Tail Workers are not yet supported for Workers with assets

)))

Aakashra oof. I'm still struggling with figuring out an appropriate practice/handling te...

A

Ashley•12/11/24, 8:17 AM

Depending how much you need to store in the session, you could always use JWTs to avoid any state

A

akashra•12/11/24, 8:18 AM

I absolutely do not want to be transferring that much data to and from a client over repeated calls or in a JWT. Also, JWTs can't easily be expunged/expired. I don't know why whenever I try to address this problem someone always insists on JWTs when they're just not an appropriate solution to a problem like this.

Aakashra I absolutely do not want to be transferring that much data to and from a client ...

I

Isaac McFadyen•12/11/24, 5:14 PM

JWTs can't easily be expunged/expired

True, but in your initial question you were concerned about writes to KV/DO/D1. If you validated the JWT on every request, it would still only be a a one-time write, and then reads after that, and if you wanted to expire it for some reason it would be 1 extra write.

I

Isaac McFadyen•12/11/24, 5:15 PM

This would also be applicable for regular sessions. You don't need to write on every request... if you're worried about the session expiring then just hold off a write until 10 minutes before expiry and then renew at that time.

M

Mozzy•12/11/24, 8:47 PM

Having some issues testing this myself, so:
Does anyone know if Hono's

getContext()

getContext()

getContext()

getContext() works inside a scheduled event triggered by a cron?
In my scheduled event, I would call a function that interacts with KV by using

getContext()

getContext()

getContext()

getContext() to get the KVNamespace

M

Mozzy•12/11/24, 8:59 PM

I assume it does not work, because there's no actual request, and therefore no context middleware

M

Mozzy•12/11/24, 9:26 PM

Is there something wrong with cron and deployments?

I've had a wrangler.toml with

crons = ["0 */1 * * *"]

crons = ["0 */1 * * *"]

crons = ["0 */1 * * *"]

crons = ["0 */1 * * *"] deployed for a long time which was working.
Earlier today, I updated that to

crons = ["* * * * *"]

crons = ["* * * * *"]

crons = ["* * * * *"]

crons = ["* * * * *"] and deployed. The dashboard UI was still saying

0 */1 * * *

0 */1 * * *

0 */1 * * *

0 */1 * * * was active, and it was still using that pattern.

Now I just reverted it and deployed

crons = ["0 */1 * * *"]

crons = ["0 */1 * * *"]

crons = ["0 */1 * * *"]

crons = ["0 */1 * * *"], and now the

* * * * *

* * * * *

* * * * *

* * * * * pattern is the one actually running, despite the dashboard UI saying

0 */1 * * *

0 */1 * * *

0 */1 * * *

0 */1 * * * is active.

M

Mozzy•12/11/24, 9:27 PM

M

Mozzy•12/11/24, 9:33 PM

I just did a deploy of the same exact content, and now it seems to be using the correct cron schedule. It seems that deployments use the previous deployment's cron triggers

SSPS | Shootmail I am deploying SvelteKit project on Workers. Auth JS with providers Google and G...

M

Mitya•12/12/24, 3:50 PM

Could be wrong but I seem to recall Workers supports certain Node features locally but not in production, an example being

eval()

eval()

eval()

eval(). This is probably another such example.

M

Mitya•12/12/24, 4:41 PM

If I need Worker A to hand over to Worker B to do some heavy lifting, without delaying the response from Worker A, are service bindings the answer here? Will Worker B complete it's job even after Worker A has sent its response (i.e. finished)?

MMitya If I need Worker A to hand over to Worker B to do some heavy lifting, without de...

A

Ashley•12/12/24, 5:31 PM

You will need to use waitUntil (https://developers.cloudflare.com/workers/runtime-apis/context/#waituntil) to make sure Worker A waits for Worker B to do its work before terminating, allowing you to return the response and then do whatever work you need in the waitUntil (could also use Queues or Workflows if you need retries / more reliability)

Cloudflare Docs

Context (ctx) · Cloudflare Workers docs

The Context API in Cloudflare Workers, including waitUntil and passThroughOnException.

AAshley You will need to use waitUntil (https://developers.cloudflare.com/workers/runtim...

M

Mitya•12/12/24, 7:47 PM

Ah but that's the point - I don't want Worker A to wait for Worker B to finish. Worker A's response has no dependency on Worker B, and I don't want it to be delayed. Worker B's work simply needs to be kicked off by Worker A and then left alone (it may take several minutes.) Given this, presumably I don't need

waitUntil

waitUntil

waitUntil

waitUntil?

MMitya Ah but that's the point - I don't want Worker A to wait for Worker B to finish. ...

A

Ashley•12/12/24, 7:54 PM

Last I heard, waitUntil gives you 30 seconds. It sounds like you want Queues or Workflows, rather than waitUntil!

M

Mitya•12/12/24, 7:56 PM

Thanks for the help but I think I'm struggling to be clear about my need. Queues are an option. But I simply need Worker A to trigger Worker B to begin a process. Worker A doesn't need to wait for Worker B to finish.

waitUntil

waitUntil

waitUntil

waitUntil suggests it would be useful only if Worker A had to wait until Worker B was comlplete, but for me it doesn't. Not sure if that makes sense!

Another way of phrasing this question is: if Worker A calls Worker B to do something, but Worker A then immediately sends a response, does that quit Worker B, or will Worker B persist in what it's doing?

D

D Trombett•12/12/24, 8:08 PM

rpc?

C

CanRau•12/12/24, 8:08 PM

sorry for crossposting

S

SuperHelpflare•12/12/24, 8:09 PM

Please do not post your question in multiple channels/post it multiple times per the rules at #

welcome-and-rules. It creates confusion for people trying to help you and doesn't get your issue or question solved any faster.

C

Cobra•12/12/24, 8:28 PM

I'm running into a problem similar to the one Bradley described (workers-and-pages-discussions) — generating PDFs is taking significantly longer (and generating significantly larger files) than expected with CF's Puppeteer, and there's currently no way to switch the headless mode to

chrome-headless-shell

chrome-headless-shell

chrome-headless-shell

chrome-headless-shell... the docs for

puppeteer.launch()

puppeteer.launch()

puppeteer.launch()

puppeteer.launch() incorrectly suggest you can pass a

PuppeteerLaunchOptions

PuppeteerLaunchOptions

PuppeteerLaunchOptions

PuppeteerLaunchOptions param, but that only works for

PuppeteerNode.launch()

PuppeteerNode.launch()

PuppeteerNode.launch()

PuppeteerNode.launch(), not

PuppeteerWorkers.launch()

PuppeteerWorkers.launch()

PuppeteerWorkers.launch()

PuppeteerWorkers.launch()

GitHub

puppeteer/docs/api/puppeteer.launch.md at main · cloudflare/puppeteer

Puppeteer Core fork that works with Cloudflare Browser Workers - cloudflare/puppeteer

MMitya Thanks for the help but I think I'm struggling to be clear about my need. Queues...

J

James•12/12/24, 8:35 PM

thats pretty much what queues are best for

C

Chill guy•12/13/24, 5:10 AM

to use cron trigger the scheduled function should be exported default in index.ts file?

C

Chill guy•12/13/24, 5:11 AM

i tried to placing the scheduled function in separate file but bot working as expected

CChill guy i tried to placing the scheduled function in separate file but bot working as ...

H

Hello, I’m Allie!•12/13/24, 5:18 AM

The actual code can be located anywhere you like, as long as it is exported in your

main

main

main

main file. For example:

// scheduled.ts
async function scheduled(evt, env, ctx) {
  // Do some work
}
// index.ts
import { scheduled } from "./scheduled";
export default { scheduled };

// scheduled.ts
async function scheduled(evt, env, ctx) {
  // Do some work
}
// index.ts
import { scheduled } from "./scheduled";
export default { scheduled };

// scheduled.ts
async function scheduled(evt, env, ctx) {
  // Do some work
}
// index.ts
import { scheduled } from "./scheduled";
export default { scheduled };

// scheduled.ts
async function scheduled(evt, env, ctx) {
  // Do some work
}
// index.ts
import { scheduled } from "./scheduled";
export default { scheduled };

C

Chill guy•12/13/24, 5:25 AM

thanks. Got it

Original message was deleted

J

James•12/13/24, 4:10 PM

Do you have

@types/node

@types/node

@types/node

@types/node installed per chance? Or is something loading

webworker

webworker

webworker

webworker or other similar types? What file openes if you browse to that erroring type?

J

James•12/13/24, 4:44 PM

Can you share a full code snippet, with how

res

res

res

res is assigned?

C

Cobra•12/13/24, 9:55 PM

can anyone offer any insights regarding this: workers-and-pages-discussions ?

U

UB•12/14/24, 2:34 AM

If anyone wants to use sourcemaps with Sentry on CF workers:
Do a dry-run to a folder (such as ./dist), do your sourcemap injection and upload on that, then deploy that dry ran file

As an example:

wrangler deploy --dry-run --outdir ./dist --minify
sentry-cli sourcemaps inject ./dist && sentry-cli sourcemaps upload ./dist
wrangler deploy --outdir ./dist --no-bundle ./dist/index.js

wrangler deploy --dry-run --outdir ./dist --minify
sentry-cli sourcemaps inject ./dist && sentry-cli sourcemaps upload ./dist
wrangler deploy --outdir ./dist --no-bundle ./dist/index.js

wrangler deploy --dry-run --outdir ./dist --minify
sentry-cli sourcemaps inject ./dist && sentry-cli sourcemaps upload ./dist
wrangler deploy --outdir ./dist --no-bundle ./dist/index.js

wrangler deploy --dry-run --outdir ./dist --minify
sentry-cli sourcemaps inject ./dist && sentry-cli sourcemaps upload ./dist
wrangler deploy --outdir ./dist --no-bundle ./dist/index.js

U

UB•12/14/24, 2:35 AM

Posting this because I spent more time than I'd like to admit on that

Why is this happening and why is no one fixing it?

Similar Threads

Similar Threads

Similar Threads