Working With the New 50 MB Query Result Limit in Cloudflare D1

Cloudflare D1 is a serverless SQL database designed for applications running on Cloudflare's platform. Because D1 is often used behind APIs, serverless functions, dashboards, and data-driven web applications, the amount of data returned by a SQL query matters as much as the query itself.

A common mistake is to think about database limits only in terms of storage size or the number of rows. Query result size can become a separate constraint.

A large query may successfully execute at the database level but still produce a response that is too large for the platform to return.

The 50 MB query result limit makes this distinction important for D1 applications. Developers should understand what the limit means, how large result sets affect application design, and how to avoid building queries that depend on returning unnecessarily large datasets.

The solution is usually not to remove the limit. Instead, applications should control how much data each request retrieves.

What Does the 50 MB Query Result Limit Mean?

The important distinction is between database storage and query result size.

An application might have a database containing hundreds of megabytes or several gigabytes of information while individual queries return only a small amount of data.

For example:

SELECT id, name, email
FROM users
WHERE active = 1;

If the query returns a reasonable number of rows, the result may be small.

But this query can become problematic:

SELECT *
FROM audit_logs;

An audit table can grow continuously. Returning every column and every row creates an increasingly large result.

The database may contain the data successfully, but the application should not assume that retrieving the entire table in one operation is appropriate.

Why Large Query Results Are a Problem

Large query results create several problems at the application level.

First, more data must be transferred between the database layer and the application.

Second, the application needs memory to process the returned result.

Third, serialization can become expensive when the result is converted into JSON.

Fourth, the HTTP response itself becomes large.

A typical API request may look like this:

Client
   |
   v
API Request
   |
   v
D1 Query
   |
   v
Large Result
   |
   v
JSON Serialization
   |
   v
HTTP Response

If the result is unnecessarily large, every stage can become more expensive.

This is why query-result limits should be treated as an application-design consideration rather than simply a database restriction.

Understand the Difference Between Rows and Result Size

A common assumption is that a query returning fewer rows is automatically safe.

That is not always true.

Consider these two queries:

SELECT id, name
FROM users
LIMIT 1000;

and:

SELECT *
FROM documents
LIMIT 1000;

Both return 1,000 rows, but the amount of data can be very different.

The second query may include large text fields, JSON documents, metadata, or other columns.

Therefore, developers should think about both:

• Number of rows.
• Size of each row.

A query returning 500 large records can potentially produce a larger result than a query returning several thousand small records.

Avoid SELECT * for Large Result Sets

One of the simplest improvements is to select only the columns the application actually needs.

Instead of:

SELECT *
FROM orders;

use:

SELECT
    id,
    order_number,
    customer_id,
    status,
    created_at
FROM orders;

This has two benefits.

The database does not need to return unnecessary columns, and the application receives a smaller result.

This is particularly important when tables contain large fields such as:

• Descriptions.
• JSON documents.
• HTML content.
• Logs.
• Metadata.
• Serialized application state.

If the API does not need a column, there is little reason to return it.

Use Pagination for Large Datasets

Pagination is one of the most practical solutions for large query results.

Instead of returning thousands of records in one request, return a manageable page.

For example:

SELECT
    id,
    name,
    email
FROM users
ORDER BY id
LIMIT 50 OFFSET 0;

The next request can retrieve another page:

SELECT
    id,
    name,
    email
FROM users
ORDER BY id
LIMIT 50 OFFSET 50;

The API can then return a response such as:

{
    "items": [
        {
            "id": 1,
            "name": "Rahul",
            "email": "[email protected]"
        }
    ],
    "page": 1,
    "pageSize": 50
}

The client does not need the entire dataset at once.

Prefer Cursor-Based Pagination for Large Tables

Offset pagination is simple, but it can become inefficient for large or frequently changing datasets.

A cursor-based approach can use the last returned ID.

For example:

SELECT
    id,
    name,
    email
FROM users
WHERE id > ?
ORDER BY id
LIMIT 50;

The API can return:

{
    "items": [
        {
            "id": 101,
            "name": "Rahul",
            "email": "[email protected]"
        }
    ],
    "nextCursor": 101
}

The next request sends the cursor back to the API.

The database can then continue from that point.

This approach is particularly useful for large tables where users may repeatedly request the next page.

Limit the Columns Returned by the Query

Pagination alone may not solve a result-size problem.

Consider:

SELECT *
FROM documents
LIMIT 100;

If every document contains a large content field, 100 records can still produce a substantial response.

A better design might be:

SELECT
    id,
    title,
    author_id,
    created_at
FROM documents
ORDER BY id
LIMIT 100;

Then provide a separate endpoint for retrieving the full document when necessary.

For example:

GET /documents
GET /documents/{id}

The list endpoint returns lightweight metadata.

The detail endpoint returns the larger content only when the user requests it.

Use Filtering Before Pagination

Pagination should usually be combined with filtering.

Instead of:

SELECT id, name, status
FROM orders
ORDER BY id
LIMIT 50;

an application may only need active orders:

SELECT id, name, status
FROM orders
WHERE status = 'active'
ORDER BY id
LIMIT 50;

This reduces the number of rows that need to be considered for the API response.

Filtering also gives the API a clearer contract.

For example:

GET /orders?status=active

is generally more useful than an endpoint that returns every order and expects the client to filter the results locally.

Avoid Returning Large Data From Internal APIs

The same principle applies to service-to-service communication.

An internal API might be tempted to return:

{
    "users": [
        ...
    ],
    "transactions": [
        ...
    ],
    "auditLogs": [
        ...
    ]
}

in one request.

This makes the API difficult to scale because consumers receive more data than they may need.

A better approach is to separate responsibilities:

GET /users
GET /transactions
GET /audit-logs

Each endpoint can apply its own filtering and pagination rules.

This also makes failures easier to handle because a large audit-log query does not have to block the retrieval of user information.

Use Aggregation When Raw Rows Are Not Needed

Sometimes an application does not actually need every record.

For example, a dashboard may need the number of orders by status.

Instead of retrieving every order:

SELECT id, status
FROM orders;

the application can ask the database to perform the aggregation:

SELECT
    status,
    COUNT(*) AS total
FROM orders
GROUP BY status;

The application receives a small result:

{
    "pending": 42,
    "completed": 185,
    "cancelled": 11
}

This is often a much better design for dashboards and reporting APIs.

The database performs the aggregation, while the application receives only the information required by the UI.

Be Careful With Large Text and JSON Columns

Large columns are a common reason for unexpectedly large query results.

Consider:

CREATE TABLE documents (
    id INTEGER PRIMARY KEY,
    title TEXT,
    content TEXT,
    metadata TEXT
);

A query like:

SELECT *
FROM documents
LIMIT 100;

could return much more data than expected.

For a list page, use:

SELECT
    id,
    title
FROM documents
ORDER BY id
LIMIT 50;

Then retrieve content only when the user opens a specific document.

This separation between list data and detail data is useful for both performance and API design.

Monitor Query Result Sizes

Applications should not wait until a production request fails to discover that a query is returning too much data.

Add observability around important database operations.

For example, application logs can record:

Query: listOrders
Rows: 50
Response bytes: 18420
Duration: 72ms

The exact logging implementation depends on the application stack.

Avoid logging sensitive data simply to measure response size. Record metadata such as row counts, duration, and response size instead.

This can help identify endpoints that gradually become larger as the underlying dataset grows.

Design API Limits Alongside D1 Limits

The database result limit should not be the only limit in the application.

For example, an API can define:

Maximum page size: 100
Default page size: 25
Maximum export size: application-specific

If the client requests:

?pageSize=100000

the API should not blindly honor the value.

Instead, cap it:

requested size = 100000
allowed size = 100

This prevents clients from accidentally creating oversized database requests.

Large Export Operations Need a Different Design

Pagination is ideal for normal API requests, but exports are different.

Suppose an administrator wants to export hundreds of thousands of records.

Trying to return the entire export through a normal API request is usually a poor design.

A better architecture is:

Export Request
      |
      v
Create Export Job
      |
      v
Process Data in Batches
      |
      v
Generate Export File
      |
      v
Store File
      |
      v
Return Download Information

The exact implementation depends on the application architecture.

The important idea is to separate large data-processing jobs from normal interactive requests.

Query Result Size vs Query Performance

These concepts are related but not identical.

A query can be fast but return too much data.

A query can also return a small amount of data but take a long time because of poor filtering or missing indexes.

For example:

SELECT id, name
FROM users
WHERE email = ?;

may return one row but still need optimization if the filtering column is not indexed appropriately.

Therefore, when improving D1 queries, evaluate both:

Common Mistakes

Using SELECT *

Returning every column makes it easy for a query to become larger as the schema evolves.

Select only the fields required by the application.

Increasing Page Size to Reduce API Calls

Increasing the page size can reduce the number of requests, but it also increases the size of each response.

Find a reasonable page size instead of assuming that fewer requests are always better.

Returning Large Documents in List APIs

List endpoints should generally return lightweight metadata.

Retrieve large content only when needed.

Ignoring Data Growth

A query that works with 10,000 rows may become a problem after the application grows.

Design pagination and filtering before the dataset becomes large.

Fetching Everything for a Dashboard

Dashboards often need counts and summaries rather than raw records.

Use database aggregation where appropriate.

Best Practices

1. Avoid SELECT * for application APIs.
2. Select only the columns required by the current operation.
3. Use pagination for large collections.
4. Prefer cursor-based pagination for large or frequently changing datasets.
5. Apply filters before returning data to the application.
6. Use aggregation for dashboards and summary views.
7. Separate list and detail endpoints.
8. Set reasonable API page-size limits.
9. Monitor query duration, row counts, and response sizes.
10. Design large exports as asynchronous jobs instead of normal API responses.
11. Review queries as database tables grow.
12. Test endpoints using realistic production-sized datasets.

Troubleshooting Large Query Results

If a D1 query begins failing because the returned result is too large, start by identifying what contributes to the result size.

Check:

• Number of rows returned.
• Number of columns selected.
• Size of text columns.
• JSON or serialized columns.
• Missing filters.
• Missing pagination.
• Unnecessary joins.
• Large nested application responses.

For example, change:

SELECT *
FROM audit_logs;

to:

SELECT
    id,
    action,
    user_id,
    created_at
FROM audit_logs
ORDER BY id
LIMIT 50;

Then retrieve detailed information separately when required.

Do not simply reduce the limit until the error disappears. The query should be redesigned around what the application actually needs.

Advantages of Designing Around the Limit

Better API Performance

Smaller responses generally require less data transfer and less application processing.

Lower Memory Usage

The application does not need to hold unnecessarily large result sets in memory.

Better User Experience

Clients receive useful data faster instead of waiting for very large responses.

Easier Scaling

Pagination and filtering prevent individual requests from growing without control.

More Predictable APIs

Explicit response limits make API behavior easier to reason about.

Disadvantages and Trade-Offs

More API Requests

Pagination means clients may need multiple requests to retrieve a complete dataset.

More Complex Client Logic

Clients need to understand pages or cursors.

Export Workflows Become More Complex

Large exports may require background processing instead of a simple request.

Additional API Design Work

List endpoints, detail endpoints, filters, cursors, and export operations require more planning.

These trade-offs are generally preferable to allowing normal requests to return unbounded amounts of data.

Summary

The 50 MB query result limit in Cloudflare D1 is a reminder that database storage and query-result size are separate concerns.

An application can store a large amount of data while still needing to keep individual query responses small.

The most effective approach is to design queries around the data the client actually needs. Avoid unnecessary columns, apply filters early, use pagination for collections, and use aggregation when the application needs summaries instead of raw records.

Large text and JSON columns deserve particular attention because a relatively small number of rows can produce a large result.

For administrative exports and other large data operations, use a separate batch or asynchronous workflow instead of trying to return everything through a normal API request.

The goal is not simply to stay below the platform limit. The goal is to build APIs that return predictable, focused, and manageable amounts of data as the application grows.