> ## Documentation Index
> Fetch the complete documentation index at: https://mixpanel-edb78807-copilot-tof-446-create-per-error-troubles.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# 429 Too Many Requests

> Mixpanel's ingestion rate limit, the exponential backoff strategy to use when you hit it, and how to run a large backfill without throttling yourself

A 429 means your project exceeded Mixpanel's ingestion rate limit. Retry with exponential backoff.

## What the Response Looks Like

```json theme={"system"}
{
  "code": 429,
  "error": "Project exceeded rate limits. Please retry the request with exponential backoff.",
  "status": "Too Many Requests"
}
```

<Note>
  Mixpanel does not return a `Retry-After` header. Compute your own backoff interval using the strategy below.
</Note>

## The Rate Limit

| Endpoint  | Limit                                                                                                         |
| --------- | ------------------------------------------------------------------------------------------------------------- |
| `/import` | 2GB of uncompressed JSON per minute, or roughly 30k events per second, measured on a rolling one-minute basis |
| `/track`  | The same limits as `/import`                                                                                  |

The limit is per project.

## How to Fix It

### Retry With Exponential Backoff and Jitter

Start with a 2-second backoff, double it up to a maximum of 60 seconds, and add 1–5 seconds of jitter so that concurrent clients do not retry in lockstep.

```python theme={"system"}
import random
import time

import requests

MAX_ATTEMPTS = 8


def send_batch(url, auth, events):
    for attempt in range(MAX_ATTEMPTS):
        response = requests.post(
            url, auth=auth, params={"strict": 1}, json=events, timeout=30
        )

        # 502 and 503 are transient too, so back off the same way.
        if response.status_code not in (429, 502, 503):
            return response

        # Start at 2s, double to a 60s ceiling, then add jitter.
        wait_time = min(2 ** (attempt + 1), 60) + random.uniform(1, 5)
        time.sleep(wait_time)

    raise RuntimeError(f"giving up after {MAX_ATTEMPTS} attempts")
```

**Do not retry a 400.** Validation errors fail consistently and still count toward your rate limit. See [400 Bad Request](/troubleshooting/errors/400-bad-request).

### Run Backfills at Full Speed Until Throttled

Rather than pacing requests with a fixed sleep, send as fast as you can with concurrent clients and let 429s tell you when to slow down. Mixpanel sees the best results with **10–20 concurrent clients sending 2,000 events per batch**.

A fixed delay between requests almost always underuses the available throughput and makes a large backfill take far longer than it needs to.

### Ask for a Higher Limit

If you are an Enterprise customer and a one-time backfill needs more headroom, contact your CSM with your `project_id` and the use case.

## Related

* [400 Bad Request](/troubleshooting/errors/400-bad-request)
* [413 Payload Too Large](/troubleshooting/errors/413-payload-too-large)
* [Import Events API reference](/reference/import-events)