Are embedded replicas planned? I mean something like turso.

DDio Are embedded replicas planned? I mean something like turso.

lambrospetrou•6/30/25, 11:27 AM

No plans at the moment. There are some other companies trying to do sync engines ontop of DOs, so if that fits your use-cases you might want to explore those. I haven't tried any personally, so cannot recommend specific ones.

Original message was deleted

lambrospetrou•6/30/25, 11:29 AM

is using 'first-primary' the equiv of using not using sessions api at all - even if i have replication enabled?

For the first call, yes, they are the same.

If you continue using the same session object though, your subsequent queries will potentially be processed by a replica that is at least up-to-date as your last read/write. If you don't use the sessions API ALL requests are going to the primary (treated as first-primaryfirst-primary).

if 'first-unconstrained' detects a write through sessions will it be upgraded to a primary, thus adding extra latency?

Yes, if a query is a write, we will automatically forward it to the primary for execution. It still flows through the replica location though, so there might be some extra latency (or not) depending on these locations.

Llambrospetrou > is using 'first-primary' the equiv of using not using sessions api at all - ev...

lambrospetrou•6/30/25, 11:31 AM

We have plans to make both the replicas placement better, and also to make the routing smarter so that if writes are detected in the query we send you straight to primary but it's still some time away from being released.

lambrospetrou•6/30/25, 12:47 PM

Hey folks, we released a small improvement last week for read replication and would love to get some feedback, or unexpected encountered issues.

Until last week, requests flowing through your replicas would wait for the replica to startup instead of using the primary until the replica is up and ready, which led to some latency increase for applications without constant traffic.
As of last week, requests now flow to the primary instance while replicas boot up, therefore latencies are expected to be better.

Keep in mind that if your application is doing less than 1 query per minute, your primary will most likely have to cold start often too, this didn't change, so in those cases you will anyway have the same latencies.

Please try it out and let us know if you encounter any issues or unexpected behavior.

Llambrospetrou Hey folks, we released a small improvement last week for read replication and wo...

Zack•6/30/25, 12:53 PM

Awesome! I will give it another go and report back sometime this week

Llambrospetrou Hey folks, we released a small improvement last week for read replication and wo...

Vero•6/30/25, 1:21 PM

@D1 Database

Llambrospetrou > is using 'first-primary' the equiv of using not using sessions api at all - ev...

ajgeiss0702•6/30/25, 3:14 PM

this reminds me.

With the http api, does that work the same as querying from a workers without a session, consistency wise?
i know the http api adds additional latency, but thats not a huge issue with what im using it for. Consistency can be though

Basically, I'm writing to the d1 database from a service running outside of cloudflare, then reading it from workers (on cf pages)

please ping me if you reply

Aajgeiss0702 this reminds me. With the http api, does that work the same as querying from a ...

lambrospetrou•6/30/25, 4:06 PM

The HTTP API does not use Sessions API for now, so all queries go to the primary.
BTW, the latency for queries has improved a lot a few weeks ago so it's not far off from the Workers Binding (still a bit slower though).

ajgeiss0702•6/30/25, 4:06 PM

got it, thanks

SnowBall•6/30/25, 11:29 PM

huh

Original message was deleted

lambrospetrou•7/1/25, 8:41 AM

Yeah, this should work. Read more at https://developers.cloudflare.com/d1/worker-api/d1-database/#batch
Let us know if you find sometihng not working as expected.

Keep in mind the first-unconstrainedfirst-unconstrained means a replica will most likely process the query.

Jumaron•7/1/25, 5:16 PM

How many queries do i need until the read replicas kick in, i still have 0 :/
Using similiar code like quankitt3 from above.
platform.env.DB.withSession(first-unconstrained);
Tried even with bookmarks but sadly nothing works.
Is there a requirement one needs to hit?

JJumaron How many queries do i need until the read replicas kick in, i still have 0 :/ Us...

Chaika•7/1/25, 5:32 PM

Do you have them enabled under Settings -> Replication?

CChaika Do you have them enabled under Settings -> Replication?

Jumaron•7/1/25, 5:33 PM

Yes

JJumaron How many queries do i need until the read replicas kick in, i still have 0 :/ Us...

lambrospetrou•7/1/25, 5:36 PM

There isn't any query limit, once it boots in the region of your requests it should start processing read requests.

Llambrospetrou There isn't any query limit, once it boots in the region of your requests it sho...

Jumaron•7/1/25, 5:36 PM

hmm, I use sveltekit maybe its because of that?

Llambrospetrou There isn't any query limit, once it boots in the region of your requests it sho...

lambrospetrou•7/1/25, 5:36 PM

DM me your account and database IDs to see if there is anything specific to it?

Jumaron•7/1/25, 5:36 PM

Sure

JJumaron hmm, I use sveltekit maybe its because of that?

lambrospetrou•7/1/25, 5:36 PM

Hmm, no idea about sveltekit.

JJumaron hmm, I use sveltekit maybe its because of that?

James•7/1/25, 9:13 PM

doubt it has anything to do w sveltekit, i use it all day, no issues

JJames doubt it has anything to do w sveltekit, i use it all day, no issues

Jumaron•7/1/25, 9:15 PM

Yes issue was resolved

Most likely the disabling and enabling the Replica button again fixed it.

Original message was deleted

lambrospetrou•7/2/25, 7:27 AM

When you don't use sessions explicitly, every request is always sent with first-primaryfirst-primary. Those examples should be identical in practice, since the replicas forward writes to the primary.

If the queries were reversed though, and the selectselect was the 2nd query happening, they would be different.

Original message was deleted

lambrospetrou•7/2/25, 1:52 PM

Nice to hear that. Let us know if you find any issue.

Ssaminmnkani @lambrospetrou

Hello, I’m Allie!•7/2/25, 6:06 PM

?pings

SuperHelpflare•7/2/25, 6:06 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.

onlyanegg•7/2/25, 9:58 PM

So, please forgive me if I'm asking an RTFM sort of question -- I've researched quite a bit, but haven't found definitive answers, I'd be happy to be pointed at the right place to look. I'm just getting started with CloudFlare Workers / D1, and I'm wanting to make sure I fully understand the security models so that I don't accidentally expose my users to security holes. I have read that cloudflare workers are serverless, which sounds to me like my full codebase (assembled) will be getting delivered to the end-users browser. If this is the case, it introduces some interesting security problems with disassembly, but I think is manageable. What I'm unclear about though is -- where are secrets stored, and how to prevent an end-user from doing something nefarious like modifying the wasm to query a different users data, for example. I'm sure something must be in place -- I'm guessing it's not really fully serverless, but I haven't been able to find out what portions are delivered to the end user, and which are not. Any pointers to where I can learn this would be greatly appreciated. Apologies for the novel

Oonlyanegg So, please forgive me if I'm asking an RTFM sort of question -- I've researched ...

onlyanegg•7/2/25, 10:07 PM

TLDR: workers aren't serverless in the way I was thinking of them. There isn't a single server, but edge servers all over the world that execute the code. I'm replying to myself because after I typed that novel, I thought of some new ways to search, and ended up at this link: https://www.cloudflare.com/learning/serverless/what-is-serverless/ - which explains it pretty well

Oonlyanegg TLDR: workers aren't serverless in the way I was thinking of them. There isn't ...

Chaika•7/2/25, 10:32 PM

lol yea it's kind of a confusing name for beginners, generally though when you see serverless anything it's just a way the provider is saying "You don't manage servers or server level issues/worries", generally platforms where you just worry about code, but there's still servers and datacenters behind the scenes

Isaac McFadyen•7/2/25, 11:08 PM

^ also note that there's something with a very similar name to Workers called "Service Workers" that are exactly what you described (code is delivered to browser and executed in browser), but they're something anyone can deploy/use and are different than Workers.

onlyanegg•7/2/25, 11:50 PM

Yeh thanks much for confirmation / clarification - I love the concept, am hoping to switch a fair bit of stuff off my own servers before too long

Jumaron•7/3/25, 11:56 AM

What is the limit of the amount of d1 databases you could increase it to? (Not the 50k, but highest limit possible by request)

dustiiin•7/3/25, 1:14 PM

Hey, how can i edit the table schema? I want to change a columns Default Value.
when i click edit table schema the only i can do is adding a new column?

Original message was deleted

dustiiin•7/3/25, 1:41 PM

Yeah it worked. But the Dashboard from D1 doesnt display it. With PRAGMA table_info('table_name'); it shows the updated default value.
Thanks to you

Gagan Suie•7/3/25, 8:23 PM

hi, how do i get a presigned url to download a file? is there a bucket.createSignedUrlbucket.createSignedUrl function?

GGagan Suie hi, how do i get a presigned url to download a file? is there a `bucket.createSi...

lambrospetrou•7/3/25, 9:29 PM

Did you mean to post this in the R2 channel?

Llambrospetrou Did you mean to post this in the R2 channel?

Gagan Suie•7/3/25, 11:19 PM

you right. sorrrryyyy

Angiewolf•7/4/25, 1:29 AM

Hi I have a little hobby website on a worker and a d1 database bound to that worker. My d1 db is only accessable after signing in to the admin user but now I see suspicious queries on my d1 and I have no idea how or what.

AAngiewolf Hi I have a little hobby website on a worker and a d1 database bound to that wor...

lambrospetrou•7/4/25, 6:19 AM

What are the suspicious queries? The 3rd one for example is probably just the listing done for the UI dashboard if I remember correctly. The counts are very low to be an abuse.

Llambrospetrou What are the suspicious queries? The 3rd one for example is probably just the li...

Angiewolf•7/4/25, 6:24 AM

Sorry, first time using d1 here. So most of these are just because I was in the d1 dashboard and it was gathering info?

Angiewolf•7/4/25, 6:25 AM

Thank you for the reply

AAngiewolf Sorry, first time using d1 here. So most of these are just because I was in the ...

lambrospetrou•7/4/25, 6:26 AM

The 1st and 3rd seem specific writes to your own tables, so I assume triggerred by you right?
The others yeah seem safe.

Angiewolf•7/4/25, 6:26 AM

1st is def me, 3rd not too sure haha

JJumaron What is the limit of the amount of d1 databases you could increase it to? (Not t...

gruntlord6•7/4/25, 8:20 PM

its a case by case basis so you won't get a public answer. If you need that many I would use durable objects instead because they don't have any hard limits in terms of amount of objects (just the amount of classes) https://developers.cloudflare.com/durable-objects/platform/limits/#how-many-durable-objects-can-i-create

Cloudflare Docs

Limits

Durable Objects are a special kind of Worker, so Workers Limits apply according to your Workers plan. In addition, Durable Objects have specific limits as listed in this page.

okHadi (parhlai.com)•7/6/25, 2:35 AM

Hey, need (any) opinion on my appraoch on developing an application on D1. I'm thinking to implement a microservices based architecture, so each logical business unit (e.g users, signups, blogs) will have their own database. This works well in my mind.

I still can't wrap my mind around the 10gb limitation on each database. Even with a microservice architecture, if I have to scale at some point, I will hit this limit with a single db. Of course cloudflare says that you can have many DBs, but how can I even make it work?

In my application, most of the data will be shared (you can think of it as a info website). So each user will almost always see the same data. So I can't do a "multi tenant" approach here.

Another option is sharding: but I'm new to DB design and still wrapping my head around it. So far, sharding doesn't fill the usercase.

This sounds more of an architecture question than a d1 question, but I'm happy to learn any advice, or any case studies that I can be pointed towards. Thanks : )

okHadi (parhlai.com)•7/6/25, 2:39 AM

follow up question: is sharding really a recommended architecture with D1?

OokHadi (parhlai.com)Hey, need (any) opinion on my appraoch on developing an application on D1. I'm t...

gruntlord6•7/6/25, 2:44 AM

You need to be more specific, what data is shared? What data isn’t shared? Why are you convinced you will hit the limit?

Ggruntlord6 You need to be more specific, what data is shared? What data isn’t shared? Why a...

okHadi (parhlai.com)•7/6/25, 2:48 AM

Well, each user can sign up, and let's say, they have a shared access to the same data, let's say blogs. The blogs can be of finance, education or any other category. For each category, I can create a database. Makes sense to me. But if I a category has so many blogs that it reaches the 10gb limit?

I'm not convinced I'll hit the limit honestly, but it is still an interesting question for me to ask. "In case" I do, what should I do to plan to scale?

Are embedded replicas planned? I mean something like turso.

Similar Threads

Similar Threads

Similar Threads