Laravel queued job triggers cURL retry error due to non-rewindable request body

The error

  GuzzleHttp\Exception\RequestException: cURL error 0: The cURL request was retried 3 times and did not succeed. … unable to rewind the body of the request …

usually means that the HTTP request’s body (most often for a POST/PUT request) is coming from a resource that isn’t rewindable by cURL. In your case the error appears only in your queued job (queue:work) environment, which suggests that the way the HTTP request is built or stored between serialization and execution might be affecting how the body is handled.

Below are steps to resolve the issue:

---

1. Inspect the Request Body

Check how the request body is built. If you’re using a stream resource (for example, from fopen or a file handle), cURL may not be able to rewind it after the first attempt. 
Solution: Convert the body to a string instead. For example, if you’re sending file contents, use file_get_contents() so that the data is a plain string.

---

2. Ensure the Job Is Not Serializing a Non-rewindable Resource

When a job is queued, any non-serializable or resource-based data is not preserved across the queue boundaries. Examine if you are binding any resource that may have been open in your HTTP request settings. 
Solution: Set up the HTTP request inside the handle method so that all resources are freshly created when the job is executed.

---

3. Review Retry Behavior

Since the error message references retries (and cURL errors are coming after the request is retried), ensure that your API call does not rely on a retryable body unless it’s fully rewindable.
Solution:
• If you’re explicitly calling Laravel’s retry method on your HTTP request chain (Http::retry(...)), consider removing it or ensuring that the request body is prepared in a way that can be resent.
• Alternatively, if you must use retries, manually rewind the body (if it’s a stream and supports rewind) before each attempt.

---

4. Environment Differences

Since it works in your local environment but fails in staging, verify that both environments use the same PHP and cURL versions/settings. Sometimes differences in cURL versions can expose issues with how streams or retries are handled.

---

Example Adjustment

Instead of using a stream for the body, you might change your code from this:

PHPCopied
// Problematic if $resource is a stream that cannot be rewound after the first attempt
$response = Http::withOptions([
    'curl' => [
        CURLOPT_FORBID_REUSE  => false,
        CURLOPT_FRESH_CONNECT => true,
    ],
])->post('https://example.com/api', [
    'body' => $resource,
]);

To something like:

PHPCopied
// Ensure the body is a string
$body = file_get_contents('/path/to/file');

$response = Http::withOptions([
    'curl' => [
        CURLOPT_FORBID_REUSE  => false,
        CURLOPT_FRESH_CONNECT => true,
    ],
])->post('https://example.com/api', [
    'body' => $body,
]);

---

Summary

The key is to make sure that the HTTP request body can be fully replayed in case of retries. Building the body as a string or recreating/reinitializing the request when the job executes in the queue (rather than capturing a resource during job creation) should solve the issue.

These steps should help you address the error you’re experiencing on your staging server when processing API calls within a queued job.