How is cache.put coordinated when two concurrent requests in the same worker attempt to update the c

How is cache.put coordinated when two concurrent requests in the same worker attempt to update the cache for the same cache key? Specifically in the case when both requests only .put after a .match failed. I’m assuming since there is no mutex or other lock mechanism that it’s basically “last request wins”.

simpson•9/18/23, 12:47 PM

We are looking to move all of our serverless functions to Cloudflare workers, one of the things we are trying to understand is how to best structure our project. Originally we wanted to have a single Worker, but with our 20-30 functions they each need their own different CRON schedules, so it seems like we'll have to make every function its own worker. However, there is a lot of shared logic between all these functions that i'd hate to duplicate many times over and have to try and manage that. Is there any best practices on how to build a large set of workers that can share libs and also have CRON flexibility?

Ssimpson We are looking to move all of our serverless functions to Cloudflare workers, on...

alotabitsOP•9/18/23, 12:47 PM

Service bindings?

Ssimpson We are looking to move all of our serverless functions to Cloudflare workers, on...

sathoro•9/18/23, 12:48 PM

you could do a single worker and instead of using crons you could use durable object alarms

Aalotabits How is cache.put coordinated when two concurrent requests in the same worker att...

sathoro•9/18/23, 12:49 PM

probably yes

sathoro•9/18/23, 12:49 PM

the cache is per colo btw

Ssathoro you could do a single worker and instead of using crons you could use durable ob...

simpson•9/18/23, 12:59 PM

interesting! so bind a single worker to a DO and put all the scheduling logic there?

Ssimpson interesting! so bind a single worker to a DO and put all the scheduling logic th...

sathoro•9/18/23, 1:00 PM

sort of, but in your case you might want multiple DOs for each cron to simplify the scheduling. although I've also seen a library that would help with this (can't find it...)

Ssimpson interesting! so bind a single worker to a DO and put all the scheduling logic th...

sathoro•9/18/23, 1:05 PM

can make your own scheduler with a single DO (but multiple instances) like this using itty-durableitty-durable

import { createDurable } from "itty-durable";

export class Scheduler extends createDurable({ autoPersist: true }) {
  constructor(state, env) {
    super(state, env);
  }

  create(name, doEverySeconds = 30) {
    this.name = name;
    this.doEverySeconds = doEverySeconds;

    this.setAlarm(Date.now());
  }

  async alarm() {
    console.log(`Doing ${this.name}...`);

    this.setAlarm(Date.now() + this.doEverySeconds * 1000);

    // do the thing...
  }
}

import { createDurable } from "itty-durable";

export class Scheduler extends createDurable({ autoPersist: true }) {
  constructor(state, env) {
    super(state, env);
  }

  create(name, doEverySeconds = 30) {
    this.name = name;
    this.doEverySeconds = doEverySeconds;

    this.setAlarm(Date.now());
  }

  async alarm() {
    console.log(`Doing ${this.name}...`);

    this.setAlarm(Date.now() + this.doEverySeconds * 1000);

    // do the thing...
  }
}

sathoro•9/18/23, 1:05 PM

and the DO can import whatever code it needs to from your codebase directly

sathoro•9/18/23, 1:06 PM

(itty-durableitty-durable is not needed, it just makes the code way shorter)

dave•9/18/23, 2:41 PM

Does .fetch.fetch on a Service Worker binding do anything with the cache-controlcache-control header?

TTraffy Does worker supports mongoose or mongodb driver?

Legendarist•9/18/23, 4:35 PM

https://www.mongodb.com/developer/products/atlas/cloudflare-worker-rest-api/

Create a REST API with Cloudflare Workers and MongoDB Atlas | MongoDB

Learn how to create a serverless REST API using Cloudflare workers and MongoDB Atlas.

LLegendarist https://www.mongodb.com/developer/products/atlas/cloudflare-worker-rest-api/

Isaac McFadyen•9/18/23, 5:06 PM

That's Atlas only but yes.

Isaac McFadyen•9/18/23, 5:06 PM

(Atlas is their hosted service)

dave•9/18/23, 6:56 PM

Can I do a Worker service binding from Worker #1 in local mode (

wrangler dev

wrangler dev

), that goes to Worker #2 in remote mode (

wrangler dev --remote

wrangler dev --remote

dave•9/18/23, 7:02 PM

https://github.com/cloudflare/cloudflare-docs/commit/b03a8368f7f0b8a9eb89a7276baef894bad71078

GitHub

fix: cannot interact with service bindings locally (#9890) · cloudf...

The documentation provides a non-existent argument for

pages dev

pages dev

. Even if it is planned to add the functionality in future, right now the instruction is rather confusing than helpful.

dave•9/18/23, 7:02 PM

seems like no?

MetalGear•9/18/23, 7:35 PM

would you guys say that cloudflare workers are fully capable of building a hotel PMS system? I have seen alot of people use workers for MVP but wondering if you guys reccomend it for a scalable saas

MMetalGear would you guys say that cloudflare workers are fully capable of building a hotel...

sathoro•9/18/23, 7:36 PM

it scales yes

sathoro•9/18/23, 7:36 PM

Web Scale

MetalGear•9/18/23, 7:37 PM

in what cases would you guys say yeah go with monolith because cloudflare workers wont be a good fit for that

Unsmart•9/18/23, 7:51 PM

I would personally never go with a monolith

Unsmart•9/18/23, 7:51 PM

theres serverless cases for probably everything

Cyb3r-Jak3•9/18/23, 7:53 PM

I mean depends on the stack. If you need lots of pieces then monolithic can be easier to maintain and deploy

CCyb3r-Jak3 I mean depends on the stack. If you need lots of pieces then monolithic can be e...

MetalGear•9/18/23, 8:17 PM

I guess I’m just wondering what the limitations of cloudflare workers are. Is it recommended to build a whole saas hotel reservation system on it for example

Unsmart•9/18/23, 8:20 PM

Can you build a hotel reservation system on workers - yes. Should you? Do you know how to use cloudflare can you hire people who are good at cloudflare.

UUnsmart Can you build a hotel reservation system on workers - yes. Should you? Do you kn...

MetalGear•9/18/23, 8:25 PM

Yes I’d say I’m good at building apis on workers with honojs. I’m a one man team

MetalGear•9/19/23, 12:05 AM

does anyone know how I can make npx wrangler dev use data the is in my deployed worker. It was using it but its now stale for some reason

MMetalGear does anyone know how I can make npx wrangler dev use data the is in my deployed ...

Cyb3r-Jak3•9/19/23, 12:12 AM

Like a binding?

CCyb3r-Jak3 Like a binding?

MetalGear•9/19/23, 12:13 AM

No like I have DO persistent storage data in my deployed worker

MetalGear•9/19/23, 12:13 AM

I want to basically run npx wrangler dev and use the latest data that is in deployed worker.

MetalGear•9/19/23, 12:14 AM

I thought that happened automatically whenever I restart npx wrangler dev but for some reason its not right now.

MetalGear•9/19/23, 12:14 AM

I am using the latest version of wrangler btw

Cyb3r-Jak3•9/19/23, 12:18 AM

Have you tried using the

--remote

--remote

flag with wrangler dev? I am not sure how it works with DO but that is the standard way for interacting with remote bindings

matt•9/19/23, 2:46 AM

Where's the best place to raise small issues with the Workers docs?

kian•9/19/23, 2:51 AM

https://github.com/cloudflare/cloudflare-docs/issues

GitHub

Issues · cloudflare/cloudflare-docs

Cloudflare’s documentation. Contribute to cloudflare/cloudflare-docs development by creating an account on GitHub.

matt•9/19/23, 2:56 AM

Thanks @kian

one other quick question, I see on the community forums it is possible to route a Worker request to a specific colo. But I can't seem to find where this is documented

kian•9/19/23, 3:05 AM

It isn’t documented since it isn’t a feature that’s available

matt•9/19/23, 3:09 AM

ok thanks! for context, im trying to build something similar to durable objects I tried DO's but performance was not great from Australia, with each round trip to DO costing 200-400ms. My target roundtrip is 20-30ms

I read that each worker has a singleton in each colo with shared memory, having access to the routing would help me build something do-like. Is that a good enough use case to gain access to custom routing?

Unsmart•9/19/23, 3:30 AM

I’m not sure they even really offer that to enterprise customers beyond maybe jurisdictions but also using global scope is something suggested not to do

matt•9/19/23, 3:32 AM

It seems to be recommended here:
https://developers.cloudflare.com/workers/platform/limits/#memory

Use global variables to persist data between requests on individual nodes

matt•9/19/23, 3:33 AM

There is also a thread about the ability to route to a specific worker DC colo: https://community.cloudflare.com/t/how-to-run-workers-on-specific-datacenter-colos/385851

matt•9/19/23, 3:35 AM

No wait, it's this one:
https://community.cloudflare.com/t/explicitly-choosing-worker-location-to-perform-request-based-on-path/383886

�

💻🐒🔫🔥🤖•9/19/23, 3:39 AM

Is it possible to make auto-deployments only trigger when a pull request is made on a github branch? As opposed to on every single commit?

Izrath•9/19/23, 3:40 AM

Im building out a discord bot which requires a response in 3 seconds. My function will take longer than that so I need to execute a separate worker separately to run my function and update discord when its done. Is it do-able to call another worker and then return/end my current worker?

kian•9/19/23, 3:42 AM

You return a deferred response and use waitUntil to patch that message

kian•9/19/23, 3:43 AM

https://developers.cloudflare.com/workers/runtime-apis/fetch-event/#waituntil

FetchEvent · Cloudflare Workers docs

In Workers, any incoming HTTP requests are referred to as "fetch" events. A Worker will respond to the HTTP request with the handler method that was …

Izrath•9/19/23, 3:56 AM

Thanks, the example I am building off of uses itty-router, which I think uses the ES modules format. This article helps, I think I was just missing the context variable. Thanks!

Mmatt ok thanks! for context, im trying to build something similar to durable objects ...

Hello, I’m Allie!•9/19/23, 7:24 AM

I would note too that even requests from the same device(and not the same app) can be routed to two different datacenters, so you aren’t guaranteed to ever hit the same one twice

HHello, I’m Allie!I would note too that even requests from the same device(and not the same app) c...

matt•9/19/23, 10:24 AM

In aware - see the community thread I linked to where it’s mentioned that you can route requests to a specific Colo after it hits the first worker

How is cache.put coordinated when two concurrent requests in the same worker attempt to update the c

Similar Threads

Similar Threads

Similar Threads