Just to be clear. D1 and SQLite DOs support Point in Time recovery for any point in time up to 30 da

Just to be clear. D1 and SQLite DOs support Point in Time recovery for any point in time up to 30 days in the past.

What I a mentioned above was for things like backups elsewhere, to different DBs, or offloading backups to another product like R2 or even different providers.

JJoão Castro There are a couple of things that look like really weird decisions to me as well...

lambrospetrouOP•11/10/25, 9:33 AM

As you will see over the next few weeks/months, a lot of these things will be improved. As a platform Cloudflare evolved a lot over the years, and yes, a lot of the mindset and priorities changed due to how customers are using the platform and how certain "magic" as you say didn't come to fruition. (Personal opinion, but I personally dislike anything "magic", if you could see it from my suggestions over the past year, so I am with you on that one

)

So, as I always say, giving constructive feedback in what you would like to see change, or even new things, and justifying with legit use-cases is the best way to steer the platform towards delivering those things.

Larry•11/10/25, 1:45 PM

These are my "rules" for working with DOs. I'm looking for feedback/pushback/thoughts. They were written for my AI Coding Agent but I believe they are good rules regardless.

Llambrospetrou Just to be clear. D1 and SQLite DOs support Point in Time recovery for any point...

plgingras88•11/10/25, 2:57 PM

Oh absolutely, PIT was already part of my answer, but thanks for clarifying! Still, PIT is only a first-level safety net and indeed doesn’t solve the real issue : offloaded backups that ensure data integrity and security and easy way to import/export. Without these "features", backups and recovery in case of emergency become far too uncertain, left entirely to the user’s discretion and prone to human error.

SQLite definitely makes things more portable than KV, though. I just hope dumping the SQL file somewhere won’t take months (or years!) to arrive.

João Castro•11/10/25, 6:55 PM

If I have a transaction going on with SQLite (Output gate closed) and inside the transaction I make a network request (fetch, Queues sendBatchsendBatch, etc), what should happen? Should it just deadlock/timeout?

JJoão Castro If I have a transaction going on with SQLite (Output gate closed) and inside the...

lambrospetrouOP•11/10/25, 7:37 PM

If I have a transaction going on with SQLite

What do you mean by this? How do you have it going? Is it an implicit transaction, or do you use the transaction()/transactionSync()transaction()/transactionSync() functions?

Llambrospetrou > If I have a transaction going on with SQLite What do you mean by this? How do ...

João Castro•11/10/25, 7:40 PM

That would be with transaction()/transactionSync()transaction()/transactionSync()

João Castro•11/10/25, 7:48 PM

For example we could have an outbox table:

Transaction starts;
Change/delete a row (Output Gate closed);
Send the message to an external system (Request blocked by the Output Gate. Deadlock?);
Receive a response;
Transaction commit/rollback depending on the response;

This should have at-least once delivery guarantee, so it would be ok to send the same message again in the case of failure to complete the transaction or SRS failing to make writes durable.

João Castro•11/10/25, 7:49 PM

Right now I think we would have to lease the message and send it outside of the transaction, and then the consumer would have to ackack it later.

JJoão Castro For example we could have an outbox table: - Transaction starts; - Change/delet...

lambrospetrouOP•11/10/25, 7:50 PM

TLDR: If the callback given to those functions throws, the transaction is rolled back. Anythinh done outside in external systems is up to you to stay consistent.

João Castro•11/10/25, 7:50 PM

So there is no way currently to make outgoing connections inside a transaction?

João Castro•11/10/25, 7:50 PM

With SQLite

JJoão Castro So there is no way currently to make outgoing connections inside a transaction?

João Castro•11/10/25, 7:55 PM

In most cases this would be a bad idea, but in the scenario I described above it should be reasonable and a simpler design

JJoão Castro That would be with `transaction()/transactionSync()`

Larry•11/11/25, 1:44 AM

I wrote this a few years back before transactionSync()transactionSync() existed but this may help: https://medium.com/cloudflare-durable-objects-design-patterns/maintaining-consistent-state-56f5bb22dba9

Medium

Maintaining consistent state

Cloudflare durable objects maintain consistency in light of storage write failure, but not when your code throws an unexpected error.

vcc•11/11/25, 1:50 AM

Can a bidirectional channel service built on the QUIC protocol ( like WebTransport) be used in durable objects?

matt•11/11/25, 2:41 AM

hey all, quick docs question

im using the cloudflare adapter with sveltekit (which works via

getPlatformProxy

getPlatformProxy

)

and this section provides instruction on using script_namescript_name in the binding to support local dev:

https://developers.cloudflare.com/workers/wrangler/api/#supported-bindings

however after setting up this way i'm seeing this error trying to call RPC methods on the stub:

Error: Cannot access "DurableObjectStore#initialize" as Durable Object RPC is not yet supported between multiple dev sessions.

Error: Cannot access "DurableObjectStore#initialize" as Durable Object RPC is not yet supported between multiple dev sessions.

the docs from the link above say this, but I'm wondering if that "not" (fourth word along) should not be there? and instead should it read "If you ARE using..."

If you are not using RPC with your Durable Object, you can run a separate Wrangler dev session alongside your framework development server.

Otherwise, you can build your application and run both Workers in the same Wrangler dev session.

Vvcc Can a bidirectional channel service built on the QUIC protocol ( like WebTranspo...

Hello, I’m Allie!•11/11/25, 3:32 AM

No, not at this tim

1984 Ford Laser•11/11/25, 4:31 AM

Are there any plans to add the agentagents Schedule API to DOs? It's one of the more useful parts of agents that I'd love to have as part of the default implementation (or if lambro or Larry wanna add it into their DO packages

)

HHello, I’m Allie!No, not at this tim

vcc•11/11/25, 5:17 AM

Does this mean that it will be integrated into this protocol in future plans ? Or can only be built using WebSocket ？

Vvcc Does this mean that it will be integrated into this protocol in future plans ? ...

Hello, I’m Allie!•11/11/25, 5:26 AM

I’m not sure. I’m not on the Durable Objects team. I only say Not at this time because I can’t guarantee that the team isn’t planning to implement it at some point

HHello, I’m Allie!I’m not sure. I’m not on the Durable Objects team. I only say Not at this time b...

vcc•11/11/25, 5:45 AM

got it , thank u !

matt•11/11/25, 5:46 AM

uploaded a very small repo showing that i cant get durable object bindings working with

getPlatformProxy

getPlatformProxy

https://github.com/mattfysh/cfdo

matt•11/11/25, 5:47 AM

following these instructions: https://developers.cloudflare.com/workers/wrangler/api/#supported-bindings

but whenever I try to invoke the DO, i just get a 503 Service Unavailable

matt•11/11/25, 5:50 AM

please take a look

really unsure how im supposed to work locally with durable objects

matt•11/11/25, 5:50 AM

is there an example repo somewhere?

matt•11/11/25, 5:56 AM

ah i see the problem -- there is no test using the wrangler_external_do.jsoncwrangler_external_do.jsonc file

https://github.com/cloudflare/workers-sdk/tree/main/fixtures/get-platform-proxy

so it probably never worked

matt•11/11/25, 5:56 AM

cool

real fun wasting a whole day on this

matt•11/11/25, 5:58 AM

in fact the only test that does use that fixture is just making sure the "internal" binding warning does not appear: https://github.com/cloudflare/workers-sdk/blob/4259256e33445b5f2ae290a68795be9c1852d860/fixtures/get-platform-proxy/tests/get-platform-proxy.env.test.ts#L286-L291

No one thought to actually then go ahead and test that the DO binding actually worked?

11984 Ford Laser Are there any plans to add the `agent`s Schedule API to DOs? It's one of the mor...

lambrospetrouOP•11/11/25, 7:49 AM

The @cloudflare/actors@cloudflare/actors library already has alarm scheduling helpers: https://github.com/cloudflare/actors/blob/9ec15ae82986adca5763ec517a9eea8d0bbe6ef7/examples/durable-objects/src/index.ts#L21

You can use it on its own inside any

DurableObject

DurableObject

, no need to extend the Actor class.

Llambrospetrou The `@cloudflare/actors` library already has alarm scheduling helpers: https://g...

1984 Ford Laser•11/11/25, 8:07 AM

That's brilliant thank you, I have a large Worker which uses agents and json-rpc-mixin which made the whole thing really heavy dependency wise. Now I can move to actors and lumenize

Kunal Verma•11/11/25, 8:13 AM

Is there a way I can know if there's durable object already present with specified name/id?
or new one needs to be initialised?

KKunal Verma Is there a way I can know if there's durable object already present with specifi...

1984 Ford Laser•11/11/25, 8:40 AM

The GraphQL API can let you filter a list of objects by DurableObjectIDDurableObjectID, using DurableObjectStorageGroups (can't remember exact name of that, but it's something like that) to be able to see which instances are using storage

1984 Ford Laser•11/11/25, 8:43 AM

AccountDurableObjectsSqlStorageGroupsAccountDurableObjectsSqlStorageGroups for SQL DOs and AccountDurableObjectsStorageGroupsAccountDurableObjectsStorageGroups for KV DOs

LLarry I wrote this a few years back before `transactionSync()` existed but this may h...

João Castro•11/11/25, 11:05 AM

The problem here is that I want to communicate with the external world while state is not durable, more specifically inside transactionSync()transactionSync()

João Castro•11/11/25, 11:06 AM

I didn't test this yet, but I'm guessing that if I try to make a fetchfetch call inside transactionSynctransactionSync it should deadlock, since it needs the response to complete the transaction, and the request needs to wait for the Output gate to be open

JJoão Castro The problem here is that I want to communicate with the external world while sta...

lambrospetrouOP•11/11/25, 11:06 AM

Communicating with outside world immediately means it's not sync though

Llambrospetrou Communicating with outside world immediately means it's not sync though 😅

João Castro•11/11/25, 11:08 AM

Yeah, I'm talking about SQLite, it could be inside a transaction()transaction() callback

JJoão Castro For example we could have an outbox table: - Transaction starts; - Change/delet...

João Castro•11/11/25, 11:08 AM

How can I achieve this with SQLite?

JJoão Castro How can I achieve this with SQLite?

lambrospetrouOP•11/11/25, 11:12 AM

Have you tried the transaction()transaction() and it throws?

Vvcc Can a bidirectional channel service built on the QUIC protocol ( like WebTranspo...

Larry•11/11/25, 12:19 PM

In order for that to be most useful from a DO, Cloudflare will also have to implement the equivalent of hibernating WebTrasport. It's on my wishlist too, but from what I understand, outside of WebRTC, Cloudflare has limited support for QUIC anywhere, so it'll be a while before that trickles all the way down to DOs.

JJoão Castro I didn't test this yet, but I'm guessing that if I try to make a `fetch` call in...

Larry•11/11/25, 12:20 PM

That sounds right but I have never used transactionSynctransactionSync myself either. It should work with transactiontransaction though.

echoes221•11/11/25, 12:50 PM

I'm doing some local development with a durable object, is there a way to force an alarm to run/trick the date of the system so it runs?

Eechoes221 I'm doing some local development with a durable object, is there a way to force ...

Larry•11/11/25, 1:09 PM

In the vitest-integration-beta, there is runDurableObjectAlarm()runDurableObjectAlarm() which will trigger your handler but not simulate time passing. I have also released https://lumenize.com/docs/testing/alarm-simulation, which will speed up time 100x and simulate alarms running. It works great so long as you avoid a thundering herd, which is quite easy to do at 100x. You can slow it down to 1x though.

Eechoes221 I'm doing some local development with a durable object, is there a way to force ...

Larry•11/11/25, 1:14 PM

Also, I'm about to release @lumenize/alarms package modeled after Actor Alarms, with a few differences including built in support for testing. Docs still need work, but can be found here

LLarry Also, I'm about to release [@lumenize/alarms](<https://github.com/lumenize/lumen...

echoes221•11/11/25, 4:03 PM

Thanks - I was looking for something that would let me force some alarms / time travel, but I ended up adding a trigger locally and updating my DB so that I can setup the right pathways, then force it to work

Eechoes221 Thanks - I was looking for something that would let me force some alarms / time ...

Larry•11/11/25, 4:11 PM

That's the right instinct! It's why I added a testing trigger to Lumenize Alarms. TBH, I rarely use my 100x speedup alarm simulation since I added that.

LLarry That's the right instinct! It's why I added a testing trigger to Lumenize Alarms...

echoes221•11/11/25, 5:00 PM

I've got a bit of a complex stripe subscription schedule setup, and needed to start a sub in future so that billing anchors etc are in the right place, so needed to trigger the future event to ensure that the alarm triggers the schedule build correctly

1984 Ford Laser•11/13/25, 1:20 AM

Do DOs still have a 1MiB limit on outgoing WS messages? The limits page states that the new 32MB message limit is only for incoming messages

Larry•11/14/25, 1:16 AM

Does anyone have any experience with setting a zero delay alarm? I assume that they won't be skipped but that's one question. However, my real question is what's the delay between when you set the alarm and when it actually fires? Is it 1-2 ms? or can it be upwards of 10ms?

LLarry Does anyone have any experience with setting a zero delay alarm? I assume that t...

lambrospetrouOP•11/14/25, 7:42 AM

It can be upwards of 10ms for sure. You are not going to get a frequency at the single-double-digit milliseconds.

Larry•11/14/25, 1:44 PM

I assume that awaiting a workers rpc call opens input/output gate? But how does not awaiting work? There might not be a fetch or message handler to finish to open them?