I went straight to DOs so I don't know much about D1, but is there a backdoor to D1's SQLite file?

LLarry I went straight to DOs so I don't know much about D1, but is there a backdoor to...

James•8/12/25, 9:24 PM

no, but some people get the impression DO sqlite is something they get more 'native' sqlite access to the file system or extensions if they don't know otherwise

red2green•8/12/25, 9:27 PM

You can connect to D1 using wrangler

red2green•8/12/25, 9:27 PM

so that's a "backdoor"

Rred2green You can connect to D1 using wrangler

LarryOP•8/12/25, 9:33 PM

I was aware of that for D1 and DO, but it sounded like James was hinting at it being available for D1 but not DO for files in the Cloudflare cloud.

red2green•8/13/25, 4:12 PM

generally speaking, other than the limits of db size, query size, etc. Are there any requests per second limits?

Rred2green generally speaking, other than the limits of db size, query size, etc. Are th...

lambrospetrou•8/13/25, 4:22 PM

Each database is backed by a Durable Object, which is single threaded SQLite under the hood, so the limit depends on how fast/slow your queries are. If your queries execute at 1ms, you can do 1K of them at most, etc.
Rough guideline is that for most cases we see SELECTs with proper indexes usually finish in less than 1ms, and writes are a few milliseconds, so depends on your case.
My ballpark generic guideline is about 300-900 queries per second.

We have users doing more, and we have users doing less.

Llambrospetrou Each database is backed by a Durable Object, which is single threaded SQLite und...

red2green•8/13/25, 4:46 PM

great thanks. ok I didn't realize D1 uses Durable objects... I really need to understand DO b/c I'm a little confused now

I thought DO were backed by sqlite, but didn't realize d1 has DO "in front" of it.

red2green•8/13/25, 4:46 PM

so I should create multiple d1's based o the rough threshold of 300-900 qps.

Rred2green great thanks. ok I didn't realize D1 uses Durable objects... I really need to...

lambrospetrou•8/13/25, 4:48 PM

D1 is essentially a layer ontop of SQLite DOs. The underlying sqlite engine is the same, since D1 uses a DO for each database.

red2green•8/13/25, 4:48 PM

got it thanks.

red2green•8/13/25, 4:49 PM

1. llms.txt 
2. prompt.txt 
3. d1 llms-full.txt

1. llms.txt 
2. prompt.txt 
3. d1 llms-full.txt

when using claude cli, which file should I ask it to read?

red2green•8/13/25, 4:49 PM

(this is the d1 doc page)

Rred2green got it thanks.

Hard@Work•8/13/25, 4:51 PM

You can probably connect it to the CF Docs MCP Server: https://docs.mcp.cloudflare.com/sse

HHard@Work You can probably connect it to the CF Docs MCP Server: https://docs.mcp.cloudfla...

red2green•8/13/25, 4:57 PM

ah ok, i'll look into that. that's probably more streamlined...

[

[A]nToNin•8/14/25, 9:37 AM

im confused on the d1 free tier does it give 5gb or 1gb storage?

[[A]nToNin im confused on the d1 free tier does it give 5gb or 1gb storage?

lambrospetrou•8/14/25, 10:19 AM

Where did you see the 1GB mentioned? https://developers.cloudflare.com/d1/platform/pricing/

darkpool•8/14/25, 6:19 PM

Architecturally would it be possible to implement Cap'n Proto as a data type for D1? Spanner recently released protocol buffer support (apparently was available for years in their internal build) and it unlocks some really cool features like type safety and 10x faster performance vs storing nested data in JSON/JSONB. Compared to Spanner's approach, Cap'n Proto seems like it would be significantly better too since it's zero-copy. Is this something that's been considered or technically feasible with D1's architecture?

Lengo•8/15/25, 4:32 AM

Is Cloudflare working on removing the 10 GB limit from D1? This feels quite limiting for a production database…

Lengo•8/15/25, 4:34 AM

Also, will D1 auto scale to any number of requests?

LLengo Is Cloudflare working on removing the 10 GB limit from D1? This feels quite limi...

lambrospetrou•8/15/25, 6:37 AM

No, 10GB is a hard limit at the moment. No immediate plans to change that.

LLengo Also, will D1 auto scale to any number of requests?

lambrospetrou•8/15/25, 6:38 AM

Also no, see my answer a few messages above for the same question: d1-database

If you need a traditional "big machine" database, you might be better off with Hyperdrive + any database provider of your choice.

Lengo•8/15/25, 6:40 AM

okay thanks

Lengo•8/15/25, 6:40 AM

I'm probably just going to use Neon... if I could provide a suggestion, a service like theirs in the Cloudflare dev platform would be very cool

Lengo•8/15/25, 6:44 AM

I'm curious... who is D1 intended for?

LLengo I'm curious... who *is* D1 intended for?

lambrospetrou•8/15/25, 6:47 AM

D1 is a more managed database over using SQLite Durable Objects directly, so it's great when you can shard your application data model across several of them. Sweet spot is a few tens so not exactly per user but not limited by the 10GB limit either, if you need more than that in the thousands you are better to use DOs directly.

Imagine apps like the trendy vibe coding platforms nowadays where you need a database for millions of independent projects and each one is self-contained and small, just as an example where it fits great. Or internal projects that fit within 10GB (or N multiples of them where sharding comes naturally) and the traffic is not crazy.

Lengo•8/15/25, 6:49 AM

Ah, so not for something like a social media app for example?

LLengo Ah, so not for something like a social media app for example?

lambrospetrou•8/15/25, 6:51 AM

Well, social media is broad. If you are building Facebook, no. If you are building your own version of a small social media platform you could involve D1 in certain aspects for sure, e.g. for things that are private to a user. But then it also depends no usage, so just by a 2-word description it's not obvious. Again, my guideline rule is that if you are fine with 0-100GB of total storage and requests are not tens of thousands per second, D1 can work.
Otherwise, if sharding is natural to your app, SQLite DOs, otherwise Hyperdrive + any external traditional database.

Lengo•8/15/25, 6:51 AM

Okay, got it. Thanks!

nejc•8/15/25, 1:04 PM

Did something change in how Cloudflare is calculating D1 DB size. All of the sudden all databases are half of the size that they were in the past. Any insights into what is happening? Thank you

Nnejc Did something change in how Cloudflare is calculating D1 DB size. All of the sud...

lambrospetrou•8/15/25, 1:53 PM

To let us check, which database size are you referring to exactly? From API responses, billing, dashboard, etc.

Llambrospetrou To let us check, which database size are you referring to exactly? From API resp...

nejc•8/15/25, 2:13 PM

I am getting the same results from API and Dashboard

Nnejc I am getting the same results from API and Dashboard

lambrospetrou•8/15/25, 2:16 PM

Which API exactly? I checked your known accounts and don't see any drop in total size from our analytics. Please DM me more details instead of publicly here.

Llambrospetrou D1 is a more managed database over using SQLite Durable Objects directly, so it'...

Mech•8/15/25, 2:22 PM

I see

Mech•8/15/25, 2:22 PM

I was under the impression that kv was for small projects

Mech•8/15/25, 2:22 PM

D1 was if you want to scale

Mech•8/15/25, 2:23 PM

But most of my data is actually in r2

Mech•8/15/25, 2:23 PM

D1 holds references only

Mech•8/15/25, 2:23 PM

So that's ok

Mech•8/15/25, 2:23 PM

Cool

ac•8/18/25, 3:26 PM

We are having tons of D1 issues this morning. Anyone else?

Aac We are having tons of D1 issues this morning. Anyone else?

lambrospetrou•8/18/25, 5:09 PM

Any details to help any investigation? account/database IDs, timeranges?
There isn't any ongoing D1 service incident. There was a transient network blip a few hours ago, but not consistent and only in some locations.

onewug•8/18/25, 10:44 PM

hi!

if i have the following monorepo:

├── packages/
│   ├── worker/  
│   └── db/ 
└── apps/   
    └── app1/
    └── app2/

├── packages/
│   ├── worker/  
│   └── db/ 
└── apps/   
    └── app1/
    └── app2/

where app1 and app2 are two separate workers with separate wrangler files that share the same db instance and generally share the same worker logic implemented in packages/worker. can i somehow point the local db to the same db? or are they forced to be generated in their own local .wrangler folder?