As always it depends on volume. If you have a few hundred requests per second then it should be fine

As always it depends on volume. If you have a few hundred requests per second then it should be fine either way.
Once you have thousands per second per DO it's where you need to start sharding more. There is no "general pattern" that works for everywhere. Different scales need different solutions. And obviously complexity scales as well.
I just mentioned that case because for example if something like Prime Video designed 1 DO per show, then at that scale it wouldn't work. If a local radio station did it, probably all is fine.

Llambrospetrou As always it depends on volume. If you have a few hundred requests per second th...

makrdev•2/12/25, 3:26 PM

Gotcha, I think I can start with the basic solution(to avoid pre-mature optimization since 500 rps is good enough for now) and then introduce sharding. Thanks

Mozzy•2/12/25, 3:39 PM

Is there any good way to gain insights into a DO's SQL data through a UI? Kind of like D1 has

MMozzy Is there any good way to gain insights into a DO's SQL data through a UI? Kind o...

Brayden•2/12/25, 6:39 PM

You can check out https://github.com/outerbase/starbasedb which has some prebuilt components that get deployed with the DO including a visual SQL & table editor. You can also connect your DO then to outerbase.com and use the tools there to visualize and modify your data too.

BBrayden You can check out https://github.com/outerbase/starbasedb which has some prebuil...

Mozzy•2/12/25, 7:50 PM

Sounds cool, I'll check it out

BBrayden You can check out https://github.com/outerbase/starbasedb which has some prebuil...

Jehu•2/12/25, 11:06 PM

Does this work with a pre-existing DO?

JJehu Does this work with a pre-existing DO?

Brayden•2/12/25, 11:15 PM

I haven’t tried but I believe you should be able to bind it to an existing DO no problems

Hhalftome I'm trying to use websockets with my DO. In the tutorials it invokes the DO via ...

Hard@Work•2/13/25, 9:51 AM

Nope, at the moment, WebSockets need to be passed via

fetch

fetch

Hhalftome In my use case I have the option of using either websockets or SSE to periodical...

Hard@Work•2/13/25, 10:19 AM

Nope, CPU Time is only when actively processing something. Just keeping a connection open uses no CPU

Hhalftome okay thanks. It will probably have to be backed by some kind of pubsub, and sinc...

Hard@Work•2/13/25, 10:53 AM

Yes, though I'm not entirely sure how Redis handles connections, it might make some traffic, even when not actively sending data

geg•2/13/25, 11:30 AM

ye u need websocket hibernation

geg•2/13/25, 11:31 AM

or make requests on an interval

HHard@Work Nope, CPU Time is only when *actively* processing something. Just keeping a conn...

Larry•2/13/25, 8:36 PM

I thought DOs were charged on clock-time, not CPU time. Regarless, @halftome, SSE is not a good choice. DOs can be stood up and torn down any time. Same applies for Workers/Page . Since the browser can't directly access DO (except through a WebSocket connection that was previously made via a Worker), you would need to keep both hops open continuously. That's inneficient from both a technical as well as financial perspective. As @egg said, you want websocket hibernation.

LLarry I thought DOs were charged on clock-time, not CPU time. Regarless, @halftome, S...

Hello, I’m Allie!•2/13/25, 8:55 PM

At least if I was understanding correctly, they wanted to terminate the SSE connection within a Worker

Hello, I’m Allie!•2/13/25, 8:55 PM

It was just if it was a webSocket that they would choose DOs

Hello, I’m Allie!•2/13/25, 8:55 PM

Might be wrong tho

BBrayden I haven’t tried but I believe you should be able to bind it to an existing DO no...

DiamondDragon•2/14/25, 6:27 AM

i'd be curious to learn more about this as im trying to get up to speed on using DOs, i had a question here before wondering how people are inspecting the data in their DO instances. Signed up for outbase to quickly check out and I see the D1 connections but not finding how to connect to a DO

1984 Ford Laser•2/14/25, 8:18 AM

What's the best way to parallelize calls to multiple DO stubs? I want a Worker to iterate over a list of

user_id

user_id

s and get a stub for each string, and read some data, and collect it together, but want to try and speed that process up as much as possible

11984 Ford Laser What's the best way to parallelize calls to multiple DO stubs? I want a Worker t...

lambrospetrouOP•2/14/25, 8:37 AM

Do parallel calls without awaiting them and wait at the end or after batches of N. https://www.npmjs.com/package/p-queue/v/6.0.2 is a popular package used for more control ovr the concurrency.

Llambrospetrou Do parallel calls without awaiting them and wait at the end or after batches of ...

1984 Ford Laser•2/14/25, 8:38 AM

exactly what I'm looking for thank you

DDiamondDragon i'd be curious to learn more about this as im trying to get up to speed on using...

Brayden•2/14/25, 12:59 PM

You need to bind your existing DO to a StarbaseDB Worker (clone this & update the wrangler.toml to point to your DO and deploy it to CF – https://github.com/outerbase/starbasedb) then you can connect to Outerbase for example with the "Starbase" option. Edit: made this how-to article if it helps https://x.com/BraydenWilmoth/status/1890408856892629313

CChaika No, but if you use SQL, the KV APIs still work, same api, and just use a hidden ...

Matt•2/14/25, 2:22 PM

That's good to know. So by enabling sqlite for the DO it no longer uses the KV and it's all sqlite? So the rules for optimizing KV with in-memory caching and skipping await for puts to rely on the output gate would no longer apply and everything is synchronous? And does usage/pricing switch over to the sqlite?

Matt•2/14/25, 2:31 PM

Well, I guess the read in-memory caching would probably still exist, but can you now use get without await?

MMatt Well, I guess the read in-memory caching would probably still exist, but can you...

lambrospetrouOP•2/14/25, 3:57 PM

The KV API is the same, so still async and I suggest keep your awaits to get good IDE intellisense, regardless how the underlying storage works.

plgingras88•2/14/25, 7:33 PM

Hey pals, I am new to DO. I have been amazed until now. Quick question : if I delete a DO instance, will it be recreated potentially elsewhere or the location is written in the stone for good ? Thank you !!

Pplgingras88 Hey pals, I am new to DO. I have been amazed until now. Quick question : if I de...

lambrospetrouOP•2/14/25, 7:50 PM

Mostly depends on the underlying Cloudflare infrastructure. It's possible to move, and they do move over periods of time. Still within the same region (location hint) though. Also, this is transparent to you and your code.

peachneo•2/14/25, 11:34 PM

are there any fleshed out examples of testing DO alarms with vitest?

Ppeachneo are there any fleshed out examples of testing DO alarms with vitest?

lambrospetrouOP•2/14/25, 11:47 PM

https://github.com/cloudflare/workers-sdk/blob/main/fixtures/vitest-pool-workers-examples/durable-objects/test/alarm.test.ts

peachneo•2/15/25, 1:31 AM

I get some strange leakage between tests...

I have a test that succeeds if other tests are commented out, or if it's anywhere but the first test to run. It will also succeed on live reload.

It fails because

const ran = await runDurableObjectAlarm(stub); expect(ran).toBe(true);

const ran = await runDurableObjectAlarm(stub); expect(ran).toBe(true);

// expects true but is false

    it("adds jobs and schedules an alarm", async () => {
      const result = await runInDurableObject(stub, async (instance, state) => {
        return await instance.addJobs([
          { url: "https://example.com" },
          { url: "https://example.org" },
        ]);
      });
      expect(result).toBe("Jobs added");

      // Run the alarm (which calls processQueue & processTimeouts)
      const ran = await runDurableObjectAlarm(stub);
      expect(ran).toBe(true);

      const status = await runInDurableObject(stub, async (instance) => {
        return await instance.getStatus();
      });
      expect(status).toEqual([{ status: "pending", count: 2 }]);
    });

    it("adds jobs and schedules an alarm", async () => {
      const result = await runInDurableObject(stub, async (instance, state) => {
        return await instance.addJobs([
          { url: "https://example.com" },
          { url: "https://example.org" },
        ]);
      });
      expect(result).toBe("Jobs added");

      // Run the alarm (which calls processQueue & processTimeouts)
      const ran = await runDurableObjectAlarm(stub);
      expect(ran).toBe(true);

      const status = await runInDurableObject(stub, async (instance) => {
        return await instance.getStatus();
      });
      expect(status).toEqual([{ status: "pending", count: 2 }]);
    });

The add jobs instance call simply inserts some records and sets an alarm

peachneo•2/15/25, 1:38 AM

I know there's this in the docs

Durable Object alarms are not reset between test runs and do not respect isolated storage. Ensure you delete or run all alarms with runDurableObjectAlarm() scheduled in each test before finishing the test.

Durable Object alarms are not reset between test runs and do not respect isolated storage. Ensure you delete or run all alarms with runDurableObjectAlarm() scheduled in each test before finishing the test.

But I believe my afterEach should cover this

  afterEach(async () => {
    // Clear any pending alarms between tests.
    await runInDurableObject(stub, async (_instance, state) => {
      await state.storage.deleteAlarm();
    });
    vi.restoreAllMocks();
  });

  afterEach(async () => {
    // Clear any pending alarms between tests.
    await runInDurableObject(stub, async (_instance, state) => {
      await state.storage.deleteAlarm();
    });
    vi.restoreAllMocks();
  });

1984 Ford Laser•2/15/25, 8:42 AM

If I have a Worker that iterates over a list of strings, gets a stub for each string, then calls

this.state.getWebSockets

this.state.getWebSockets

on each stub, is this 1 DO request per stub? Do these count toward the subrequest limit of the initially invoked Worker?

sneakyf1shy•2/15/25, 3:38 PM

is there a proper way to use durable objects inside NextJS on cloudflare pages?

sneakyf1shy•2/15/25, 3:38 PM

I can't seem to find anything new online, only like a github issue closed in 2023

Ppeachneo I get some strange leakage between tests... I have a test that succeeds if othe...

Fernando Barroso•2/15/25, 5:13 PM

I had something similar happen to me and the way I was able to fix it was by trying different combinations of isolated storage and single worker configs that made sense for my type of tests:

https://developers.cloudflare.com/workers/testing/vitest-integration/isolation-and-concurrency/

I also found that I was placing these params on the wrong object in my vitest config file so these params weren’t working until I fixed that (which took me freaking days to figure out it was so frustrating

Cloudflare Docs

Isolation and concurrency · Cloudflare Workers docs

Review how the Workers Vitest integration runs your tests, how it isolates tests from each other, and how it imports modules.

Ppeachneo I know there's this in the docs ``` Durable Object alarms are not reset between...

Fernando Barroso•2/15/25, 5:14 PM

Also you may want to just run the alarm in your afterEach instead of deleting using runInDurableObject

Fernando Barroso•2/15/25, 5:14 PM

That also helped me

iLearn•2/16/25, 10:12 AM

DO-Compute Billing Time Vs Workers Compute Billing Time

AIM : To understand DO Compute billing.
(PLEASE NOTE : i am aware that DO billing is based on Wall Clock time while Workers billing is based on CPU time)

Simple Use-case : Make an entry in Workers-KV (key : "hello", value "world" )

IF, Workers consumes 10ms (and hence billing time = 10ms) for above simple Use-case, is it "FAIRLY ACCURATE" to assume that the DO will also consume 10ms (and hence billing time = 10ms) for the same Use-case.

Will the DO billing charge for cold-starts etc ?

IiLearn **DO-Compute Billing Time Vs Workers Compute Billing Time** AIM : To understand...

lambrospetrouOP•2/16/25, 10:48 AM

Will the DO billing charge for cold-starts etc ?

No, you pay only the time after your DO is running.

IF, Workers consumes 10ms (and hence billing time = 10ms) for above simple Use-case, is it "FAIRLY ACCURATE" to assume that the DO will also consume 10ms (and hence billing time = 10ms) for the same Use-case.

Assuming that the

10ms

10ms

is actual round trip time for your operation, then yes, DOs should have similar performance as Workers. It's the same runtime so code behavior is identical.

One note, a DO is in a single location though, whereas a Worker can be anywhere or even smart placed. So, latencies might be different from Workers just because of where your DO will be located.

Llambrospetrou > Will the DO billing charge for cold-starts etc ? No, you pay only the time aft...

sneakyf1shy•2/16/25, 2:53 PM

heya - is there any docs or example/way to use

@cloudflare/next-on-pages

@cloudflare/next-on-pages

with DO's?

Llambrospetrou > Will the DO billing charge for cold-starts etc ? No, you pay only the time aft...

iLearn•2/16/25, 5:02 PM

Thanks for crisp answer.

lambrospetrouOP•2/16/25, 5:26 PM

Hey folks, I just released

durable-utils@0.3.0

durable-utils@0.3.0

: https://github.com/lambrospetrou/durable-utils#readme

The main update is the

StaticShardedDO

StaticShardedDO

abstraction extended with more methods

one/some/all()

one/some/all()

and

tryOne/trySome/tryAll()

tryOne/trySome/tryAll()

. There are a couple other utilities added for retries like

tryN(...)

tryN(...)

.

If anyone is sharding their DOs, try out the

StaticShardedDO

StaticShardedDO

class and let me know what you think or what's not convenient.

Read the docs first to avoid any data loss. This is "static number of shards". You cannot change them after creation.

GitHub

GitHub - lambrospetrou/durable-utils: Utilities for Cloudflare Dura...

Utilities for Cloudflare Durable Objects and Workers - lambrospetrou/durable-utils

iLearn•2/16/25, 5:41 PM

DO-Compute Cost Vs Workers Compute Cost
(Adding Update to avoid any confusion :
Assumption : Both, DO and Workers are NOT performing any IO operations)

Since, both costing are based on different metric, so done a simple back-of-the paper mathematics to have common metric to compare.

Worker Compute Pricing (after FREE Tier) : $0.02 per million CPU milliseconds
= (0.02 ) / (1,000,000 * 128MB * 0.001s) = (0.02) / (1000 * 128 MB-s) = (0.02) / (128 GB-s)
= $ 0.00015625 per GB-second

DO Compute Pricing (after FREE Tier) : $12.50/million GB-s = $ 0.0000125 per GB-second

Conclusion : Workers Compute is 12.5x times COSTLIER than DO-Compute. (Subject to the Assumption stated above)
(Though, funny part is DO necessarily requires a parent Worker for itself to be invoked, so using DO necessarily requires 2 compute - Parent Worker Compute + DO Compute)

Please confirm /update above calculation and understanding.

it("adds jobs and schedules an alarm", async () => { const result = await runInDurableObject(stub, async (instance, state) => { return await instance.addJobs([ { url: "https://example.com" }, { url: "https://example.org" }, ]); }); expect(result).toBe("Jobs added"); // Run the alarm (which calls processQueue & processTimeouts) const ran = await runDurableObjectAlarm(stub); expect(ran).toBe(true); const status = await runInDurableObject(stub, async (instance) => { return await instance.getStatus(); }); expect(status).toEqual([{ status: "pending", count: 2 }]); });

Durable Object alarms are not reset between test runs and do not respect isolated storage. Ensure you delete or run all alarms with runDurableObjectAlarm() scheduled in each test before finishing the test.

afterEach(async () => { // Clear any pending alarms between tests. await runInDurableObject(stub, async (_instance, state) => { await state.storage.deleteAlarm(); }); vi.restoreAllMocks(); });

As always it depends on volume. If you have a few hundred requests per second then it should be fine

Similar Threads

Similar Threads

Similar Threads