If your users upload code to your service and you run the code for them, you need to use Workers for

If your users upload code to your service and you run the code for them, you need to use Workers for Platforms, otherwise this is insecure. Otherwise, it's up to you.

RRahul gangotri Right, i dont have money to buy enterprise plan tho 😜

kentonOP•1/22/25, 8:38 PM

for what it's worth, this is a rumor / conspiracy theory. We do not force people to upgrade to enterprise plans just because they "got big" or whatever.

Kkenton for what it's worth, this is a rumor / conspiracy theory. We do not force people...

geg•1/22/25, 8:39 PM

May I ask what you think of my setup I described above? Is it okay to do that to save costs?

Kkenton If your users upload code to your service and you run the code for them, you nee...

Rahul gangotri•1/22/25, 8:40 PM

I dont want user to upload their code rather manage channel subscriptions and all myself, just 1 worker be enough? With multiple DO per app for my user?

Rahul gangotri•1/22/25, 8:40 PM

Similar to ably or pusher

Ggeg May I ask what you think of my setup I described above? Is it okay to do that to...

kentonOP•1/22/25, 8:40 PM

it sounds fine to me, if your DO is not getting overloaded

RRahul gangotri I dont want user to upload their code rather manage channel subscriptions and al...

kentonOP•1/22/25, 8:41 PM

If it it works, it works. I don't know enough details about your app to say if you will have scalability issues or something.

Kkenton it sounds fine to me, if your DO is not getting overloaded

geg•1/22/25, 8:41 PM

Cool!! Yep it definitely doesn't get overloaded this way!

Kkenton If it it works, it works. I don't know enough details about your app to say if y...

Rahul gangotri•1/22/25, 8:42 PM

I am telling its a pusher like service saas

Rahul gangotri•1/22/25, 8:42 PM

You know pusher right?

kentonOP•1/22/25, 8:42 PM

I don't

Ggeg Cool!! Yep it definitely doesn't get overloaded this way!

geg•1/22/25, 8:44 PM

It would be super epic if this was a native feature. But this works for now

Kkenton I don't

Rahul gangotri•1/22/25, 8:44 PM

Ok no problem, so there is no limit on websocket connection right?

RRahul gangotri Ok no problem, so there is no limit on websocket connection right?

kentonOP•1/22/25, 8:45 PM

there's just the memory limit and the fact that the DO is single-threaded. Each new connection presumably takes some memory and some CPU time to handle messages, but there's no explciit limit on number of connections.

Rahul gangotri•1/22/25, 8:46 PM

Btw here

https://pusher.com/

Pusher | Leader In Realtime Technologies

Simple, scalable and reliable. Hosted realtime APIs loved by developers and trusted by giants. Build live dashboards, notifications, geotracking, chat and more.

Kkenton there's just the memory limit and the fact that the DO is single-threaded. Each ...

Rahul gangotri•1/22/25, 8:46 PM

And whats the memory limit?

geg•1/22/25, 8:47 PM

We can read the docs for this!

Rahul gangotri•1/22/25, 8:48 PM

No need of alarms or ping pong when hibernation is used

geg•1/22/25, 8:49 PM

I encourage everyone to read docs! We can get more intricate responses from staff this way!

geg•1/22/25, 8:49 PM

That's all, bye

TTyveris Would love a quick sanity check on my planned architecture. Worker + DO w/ WS p...

Rahul gangotri•1/22/25, 8:53 PM

As mentioned in docs, You can create as many objects as needed, only limit is on namespaces per account

You can create unlimited durable objects in 1 namespace

Also there is a storage limit in each DO AND ACCOUNT LEVEL BOTH

BBahaa Is there a particular reason you find them not easy? My understanding is `wrangl...

Hard@Work•1/22/25, 9:47 PM

SvelteKit, at least afaik, doesn't have a way of allowing you to export a DO class in build, so you have to inject it manually

Hard@Work•1/22/25, 9:47 PM

In addition, the way SvelteKit does bindings when using their dev service also doesn't allow you to export a DO Class, which means you need a separate

wrangler dev

wrangler dev

process for it

James•1/22/25, 9:49 PM

I didn't know about the getPlatformProxy approach, but that would work too, I ended routing all svelkite DO requests to the local fetch handler of the DO whenever running locally, otherwise use the DO binding directly. Both return the same response signatures.

森

森@VerseLink (RW)•1/23/25, 12:39 AM

I’ve been thinking, the total account level limit for DO is 50GB, but D1 has limit of 250 GB; it was reasonable back when DO isn’t meant for database storage or as a self contained database, but now, using it as some sort of distributed database with great potential to be low latency and as near as the user as possible, does cloudflare consider:

Raising the limitations of per account DO storage
Any form of guarantee that we can always raise of said limitations; filling out this form to raise limitations seems like a weak promise, that we felt like if we do ever hit those limitations, we need to look for plan B as soon as possible in case no one replies to our limitation increase request
DO was more or less meant for “session” like behavior prior to SQLite backend, but now its also work as a distributed database, correct me if I’m wrong, but once a DO is created it cannot move location, are there plans to say either replicate the DO or hint it to initiate said instance to a different location next time it gets reinitiated (let’s say a user that has a DO db assigned to them travels between countries)
We have platform for workers, but what about platforms for DOs? It would definitely be an interesting aspect to create services on top of DO too and not just workers.

森森@VerseLink (RW)I’ve been thinking, the total account level limit for DO is 50GB, but D1 has lim...

James•1/23/25, 1:47 AM

You bring up some good points. I’d always assumed DO account level storage limits would scale up by a huge factor once the service is GA. Part of the reason I moved most of my storage to DO was to get around D1 limits for storage. 10Gb isn’t enough, and if I’m scaling horizontally, it made more sense to move to DO sql db tenants.

JJames You bring up some good points. I’d always assumed DO account level storage limit...

森

森@VerseLink (RW)•1/23/25, 2:09 AM

Agreed, I started to explore DO sql because of D1 limits (and also explored turso), but when I understood the potential of DO with SQL, I really want to work some new projects on it, my biggest concern now has been that DOs artificial limitations made sense with old DO, but with what it could do with SQL, I think the restrictions needs to be heavily revised

HowlOftheSun•1/23/25, 4:39 AM

Is that code in a shareable place? I'd be interested in seeing what you did

森森@VerseLink (RW)I’ve been thinking, the total account level limit for DO is 50GB, but D1 has lim...

lambrospetrou•1/23/25, 8:26 AM

The account storage limit for DOs is purely a soft limit. If you ask for a raise you will get a raise, assuming you are not doing something obviously phishy on the platform. I am guessing the team will raise it for everyone at some point in the future too, but for now you can ask for a raise of it once you are close to it.

Aalgads Pleasantly surprised this evening. I wanted to see if the outbox pattern could ...

lambrospetrou•1/23/25, 8:33 AM

Is there any documentation about what features in sqlite are available within the DO context? Is it safe to assume all the big ones like views, triggers, etc.?

Unfortunately, the docs are a bit spread now (we are fixing them gradually), but you can see D1's SQL docs and they are the same for DOs too.

https://developers.cloudflare.com/d1/sql-api/sql-statements/ (pragmas you can use)
https://developers.cloudflare.com/d1/sql-api/foreign-keys/ (foreign key relevant things)
https://developers.cloudflare.com/d1/sql-api/query-json/ JSON

As of today DOs use SQLite 3.47.

Aalgads Pleasantly surprised this evening. I wanted to see if the outbox pattern could ...

森

森@VerseLink (RW)•1/23/25, 9:38 AM

here's a another question: (maybe similarly applying to D1 too), how is trigger execution priced? is it priced the same rate as DO or likely no charge at all

森

森@VerseLink (RW)•1/23/25, 9:39 AM

(also ignore the question i just deleted, i brain farted)

森森@VerseLink (RW)here's a another question: (maybe similarly applying to D1 too), how is trigger ...

lambrospetrou•1/23/25, 10:08 AM

There isn't separate pricing. As long as your DO runs, you are charged as usual.

arbfay•1/23/25, 11:46 AM

Flash DX improvement idea:

instead of

const doid = env.MY_DO.idFromString(user_id)
const stub = env.MY_DO.get(doid)

const doid = env.MY_DO.idFromString(user_id)
const stub = env.MY_DO.get(doid)

let stubs be retrieved with just a string (to reduce boilerplate)

const stub = env.MY_DO.get(user_id) // ID
// or, for names
const stub = env.MY_DO.get(user_id, "name") // like KV's type

const stub = env.MY_DO.get(user_id) // ID
// or, for names
const stub = env.MY_DO.get(user_id, "name") // like KV's type

just one less line to work with DOs

Aarbfay Flash DX improvement idea: instead of ```js const doid = env.MY_DO.idFromStrin...

Max (@rozenmd)•1/23/25, 12:15 PM

When you get a network disconnect and want to recreate the stub, you don't want to recompute the ID though

elunesarrow•1/23/25, 12:19 PM

Hi, let's say I'm using DO only to rpc into its sqlite data, no websocket session, do I get charge for anything outside rpc (i.e. startup time & idle time post-rpc if any?)

MMax (@rozenmd)When you get a network disconnect and want to recreate the stub, you don't want ...

arbfay•1/23/25, 12:20 PM

is it an expensive operation? and why should that be the developer's concern and not the binding/platform's concern?

If your users upload code to your service and you run the code for them, you need to use Workers for

Similar Threads