Is provisioning a db (primary, non replicated) for each user a stupid idea? Is D1 designed for this?

Is provisioning a db (primary, non replicated) for each user a stupid idea? Is D1 designed for this?
The database size limit would make sense with this model as 1gb is plenty for each user (but not enough for multiple).
Latency would also be fantastic while keeping consistency intact.

Ccyan Is provisioning a db (primary, non replicated) for each user a stupid idea? Is D...

Dahoom•8/8/23, 10:23 AM

Theres currently a limit of 10 dbs, https://developers.cloudflare.com/d1/platform/limits/
I would say create a table for each user but that aswell might not be the best idea, since each table does take extra space as skye said

Limits · Cloudflare D1 docs

Many of these limits will increase during D1’s public alpha. Join the #d1-open-alpha channel in the Cloudflare Developer Discord to keep up to date …

Dahoom•8/8/23, 10:25 AM

maybe talk with decide.gg to help design a database schema for your use case, thats what i do

rotator•8/8/23, 10:26 AM

Hey, I'm trying to compare pricing of D1 vs Durable Objects Transactional Storage API (on the $5 plan):

D1: 25,000 mil 4 KB reads included = $0
DO: 1 mil 4KB reads included, then (24,999 mil * $0.2/mil) = $4,999

That's just the free allowance, the base price for reads is also 200x. Writes are the same base price, but the free allowance of D1 alone will cost 49$ for Durable Objects. Am I missing something here? My use case could fit both, but the pricing is encouraging me to use Workers+D1 or replace the storage of DO with D1.

DDahoom Theres currently a limit of 10 dbs, https://developers.cloudflare.com/d1/platfor...

cyanOP•8/8/23, 10:27 AM

thanks!

Rrotator Hey, I'm trying to compare pricing of D1 vs Durable Objects Transactional Storag...

Dahoom•8/8/23, 10:31 AM

I would recommend d1 but its still in alpha and there are current caps/limitations, to me the best thing is its just sql so easy to intigrate and build with

Dahoom•8/8/23, 12:03 PM

Does anyone have an example worker/function using the query JSON api? I'm having a hard time getting a grasp of it from the docs

Dahoom•8/8/23, 12:23 PM

anything really

Dahoom•8/8/23, 12:24 PM

but prepared would be preferrable

DDahoom I would recommend d1 but its still in alpha and there are current caps/limitatio...

chientrm.com•8/8/23, 12:24 PM

D1 is based on Sqlite so that migration is not really an issue.

Original message was deleted

Cyb3r-Jak3•8/8/23, 12:28 PM

There is not. D1 is primarily for workers.

Original message was deleted

chientrm.com•8/8/23, 12:29 PM

You can use this package to run a D1 proxy and using D1 from your PHP https://www.npmjs.com/package/cf-workers-proxy

npm

cf-workers-proxy

Enable Cloudflare Workers runtime for local development. Latest version: 0.3.0, last published: 17 days ago. Start using cf-workers-proxy in your project by running

npm i cf-workers-proxy

npm i cf-workers-proxy

. There are no other projects in the npm registry using cf-workers-proxy.

chientrm.com•8/8/23, 12:30 PM

nvm the package is js-only

Dahoom•8/8/23, 12:31 PM

you can do a makeshift API with cloudflare workers and call it within php

Dahoom•8/8/23, 12:32 PM

since the worker will be connected to d1 you could set a secret env variable with an API key, authenticate through the bearer header and pass in sql strings to be executed

Dahoom•8/8/23, 12:35 PM

I've done similar but it was in php for cloudflare pages, as long as you write the token validation correctly everything should be smooth

Dahoom•8/8/23, 12:36 PM

I've never tried firebase but probably?

Original message was deleted

Cyb3r-Jak3•8/8/23, 12:44 PM

The D1 API is not public and if it does become public it will fall under the API rate limit. Making a worker to proxy is your best bet

Ccyan Is provisioning a db (primary, non replicated) for each user a stupid idea? Is D...

GuillaumeLakano•8/8/23, 12:46 PM

It's not a stupid idea, and it's add a strong security layer, it's currently what we doing with DO ( a mini DB for each user ) and we wait about the same possibility in D1
D1 is based on DO, DO have unlimited Objects, and D1 will increase from 10 databases per account to « thousands to tens of thousands of tenants » , but it's still not enough if you have (expect) more users than this limit.

EDIT: I've said « D1 is based on DO » because I've read that here before, but Matt add this precisions: d1-database

GGuillaumeLakano It's not a stupid idea, and it's add a strong security layer, it's currently wha...

Matt Silverlock•8/8/23, 1:40 PM

We haven’t communicated any upper bound. We just have limits during alpha.

MMatt Silverlock We haven’t communicated any upper bound. We just have limits during alpha.

GuillaumeLakano•8/8/23, 2:00 PM

So, may be I've misunderstood you ( « We have some ideas here for the “thousands to tens of thousands of tenants” - that will isolate each DB you create. Can’t share more yet and won’t be “next month”. » in d1-database )

GuillaumeLakano•8/8/23, 2:02 PM

could be really useful if CF can officially claims this will be unlimited, so this means we can start new project, using D1 directly ( while it's in development period, until the stable release of D1 one day )

GGuillaumeLakano could be really useful if CF can officially claims this will be unlimited, so th...

Matt Silverlock•8/8/23, 2:40 PM

We at least anticipate “tens of thousands” by the end of the year. I’m not in the habit of making claims that we haven’t tested but there are no architectural reasons why millions will not work.

GGuillaumeLakano could be really useful if CF can officially claims this will be unlimited, so th...

Matt Silverlock•8/8/23, 2:41 PM

If it’s easier I can just say “eventually unlimited”. If you have millions of users we will make it work. One of the reasons we built Workers for Platforms.

GuillaumeLakano•8/8/23, 2:59 PM

Awesome, thank you @Matt for theses precisions!

rotator•8/8/23, 3:41 PM

Any comment regarding this? d1-database
To rephrase: if D1 is built on top of DO, then why is it so much cheaper?

Interestingly enough, I can wrap my D1 calls into DO transactions like that and it seems to work fine

    const value = 0;
    let d1Response: any;
    await this.state.storage.transaction(async (txn) => {
      try {
        txn.put('value', this.value);
        const stmt = this.db
          .prepare(`INSERT INTO items (itemData) VALUES (?) RETURNING *`)
          .bind(JSON.stringify({ value }));
        d1Response = await stmt.all();
      } catch (e) {
        console.error(e);
        txn.rollback();
        throw e;
      }
    });

    const value = 0;
    let d1Response: any;
    await this.state.storage.transaction(async (txn) => {
      try {
        txn.put('value', this.value);
        const stmt = this.db
          .prepare(`INSERT INTO items (itemData) VALUES (?) RETURNING *`)
          .bind(JSON.stringify({ value }));
        d1Response = await stmt.all();
      } catch (e) {
        console.error(e);
        txn.rollback();
        throw e;
      }
    });

but I wouldn't want to do this just for cost management purposes

Rrotator Any comment regarding this? https://discord.com/channels/595317990191398933/9920...

Matt Silverlock•8/8/23, 3:56 PM

I've never said it's built on top of DO. How they store data is entirely distinct.

marko•8/8/23, 4:17 PM

I'm testing D1 and I followed the comments API example but I'm getting

no such table

no such table

error.

wrangler d1 info

wrangler d1 info

shows table count correctly. Any ideas what I'm doing wrong?

MMatt Silverlock I've never said it's built on top of DO. How they store data is entirely distinc...

rotator•8/8/23, 4:22 PM

I see, thanks for clarifying this.

I can infer that for overlapping use cases D1 is heavily encouraged, unless prices change before D1 is generally available.

Mmarko I'm testing D1 and I followed the comments API example but I'm getting `no such ...

marko•8/8/23, 4:32 PM

Just noticed that

wrangler dev --remote

wrangler dev --remote

works, so I'm guessing that this is related to local env.

marko•8/8/23, 4:36 PM

But now with --remote the worker doesn't connect via websocket, uhm.

GGuillaumeLakano It's not a stupid idea, and it's add a strong security layer, it's currently wha...

cyanOP•8/8/23, 6:19 PM

That’s great thanks!

Rrotator Any comment regarding this? https://discord.com/channels/595317990191398933/9920...

Dahoom•8/8/23, 6:26 PM

How did you get syntax highlighting on that example code?

DDahoom How did you get syntax highlighting on that example code?

cyanOP•8/8/23, 6:27 PM

https://gist.github.com/matthewzring/9f7bbfd102003963f9be7dbcf7d40e51#syntax-highlighting

Gist

A guide to Markdown on Discord.

A guide to Markdown on Discord. GitHub Gist: instantly share code, notes, and snippets.

cyanOP•8/8/23, 6:27 PM

its a discord syntax highlighting block

Dahoom•8/8/23, 6:28 PM

test test

test test

Dahoom•8/8/23, 6:28 PM

interesting, I didnt know highlightjs comes with discord

MMatt Silverlock I've never said it's built on top of DO. How they store data is entirely distinc...

Adi•8/8/23, 9:43 PM

Follow up question on this. I remember reading somewhere that DO will soon have an sqlite store connected as part of it's storage layer. Would this be using the same backend as D1?

sty•8/8/23, 10:06 PM

is kv still faster than d1 for query? or is d1 catching up/on par now?

Ggeelen Of course, you _can_ use the Cache API yourself (https://developers.cloudflare.c...

Chaika•8/8/23, 11:33 PM

unless I'm missing something or something changed recently, you couldn't use Cache API to match KV's new architecture, since KV 2.0 is something more close to Tiered Caching with generic + regional tier (which is Enterprise), and the Cache API doesn't interact with tiered caching at all.
KV would be faster over D1 if eventual consistency is fine/if you can have data cached for a while (and your data fits it), especially with its recent improvements.
(It would be cool if Cache API got Tiered caching one day

)

CChaika unless I'm missing something or something changed recently, you couldn't use Cac...

Codename_404•8/9/23, 2:11 AM

It would be great if they allowed caching with r2 custom domain. Then we could use fetch with custom domain on r2 with worker. Which would enable tired caching by default.

The r2 domain could be as part of a zone, just like worker route.

Would make things much easier.

Greg•8/9/23, 6:20 AM

Has anybody used the kysely adapter for d1 and run into this error?

Do not know how to serialize a BigInt

Do not know how to serialize a BigInt

more info here https://github.com/aidenwallis/kysely-d1/issues/21

Greg•8/9/23, 6:45 AM

Also unrelated, I'm created a table with this query:

CREATE TABLE Casts (
    id int8 NOT NULL PRIMARY KEY,
    created_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
    updated_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
    deleted_at timestamp,
    timestamp timestamp NOT NULL,
    fid int8 NOT NULL,
    hash bytea NOT NULL,
    parent_hash bytea,
    parent_fid int8,
    parent_url text,
    text text NOT NULL,
    embeds text NOT NULL DEFAULT '{}',
    mentions int8 NOT NULL DEFAULT '{}',
    mentions_positions int2 NOT NULL DEFAULT '{}'
);

CREATE TABLE Casts (
    id int8 NOT NULL PRIMARY KEY,
    created_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
    updated_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
    deleted_at timestamp,
    timestamp timestamp NOT NULL,
    fid int8 NOT NULL,
    hash bytea NOT NULL,
    parent_hash bytea,
    parent_fid int8,
    parent_url text,
    text text NOT NULL,
    embeds text NOT NULL DEFAULT '{}',
    mentions int8 NOT NULL DEFAULT '{}',
    mentions_positions int2 NOT NULL DEFAULT '{}'
);

I know it has some less common data types but the table seems to create just fine and I can add data from the Cloudflare dashboard.

Only problem is when I try to read it from a worker, I always gets

D1_ERROR: SqliteError: no such table: Casts

D1_ERROR: SqliteError: no such table: Casts

akazwz•8/9/23, 6:57 AM

look like than cloudflare pages with d1 binding has some problems, my account id is a275d70ec37a6046e6778c77d65dd38b, i am using remix with pages, the DB binding will be D1_BETADB and do not work.

Aakazwz look like than cloudflare pages with d1 binding has some problems, my account id...

forslunds•8/9/23, 10:20 AM

im using Remix/Pages/D1 without problems, atleast on Dev/Local. Were are you struggling? Get any errors?

Hasi•8/9/23, 10:41 AM

I'm building a remix app

CREATE TABLE Casts ( id int8 NOT NULL PRIMARY KEY, created_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP, updated_at timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP, deleted_at timestamp, timestamp timestamp NOT NULL, fid int8 NOT NULL, hash bytea NOT NULL, parent_hash bytea, parent_fid int8, parent_url text, text text NOT NULL, embeds text NOT NULL DEFAULT '{}', mentions int8 NOT NULL DEFAULT '{}', mentions_positions int2 NOT NULL DEFAULT '{}' );

Is provisioning a db (primary, non replicated) for each user a stupid idea? Is D1 designed for this?

Similar Threads

Similar Threads

Similar Threads