How to resume a download with cURL

Large downloads can fail due to interruptions, wasting time and bandwidth if restarted from scratch. cURL supports resuming if the server allows HTTP Range requests, enabling download continuation from the last received byte.

Large file transfers often fail because of unstable connections, power issues, or temporary server problems, and restarting from the beginning wastes both time and bandwidth. Resuming an interrupted transfer with cURL continues from the last completed portion instead of re-downloading data that has already been saved locally. This approach is especially useful when working with multi‑gigabyte images, archives, or backups over long‑distance links.

Download resumption relies on server support for HTTP Range requests, where the client asks for bytes starting at a specific offset in the file. cURL provides the --continue-at (short form ‑C) option to resume based on the existing file size, sending a range request and appending the remaining data to the partial file. When the server responds with partial content, the result is a single, complete file made from both the original and resumed transfer.

Not every server accepts range requests or keeps the underlying file stable across interruptions, and resuming against a changed or corrupted partial file can generate invalid output. A partial file must remain on disk with the same name used for the original download, and integrity checks such as SHA‑256 hashes or trusted file size references help confirm that the resumed result is correct before use in production workflows.

1. Open a terminal in an environment where cURL is installed and available on the command path.

1. Start a new download of a large file using cURL and --remote-name so the local filename is derived from the URL.

   $ curl --remote-name "https://www.example.com/largefile.zip"
     % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                    Dload  Upload   Total   Spent    Left  Speed
     43  102M   43 44.1M    0     0  2130k      0  0:00:49  0:00:21  0:00:28 2280k
   ##### snipped #####

   The progress meter demonstrates that bytes are being written to the file largefile.zip on local storage.

1. Interrupt the transfer so that a partial file remains on disk for later resumption.

   Stopping the transfer with Ctrl+C cancels the HTTP request but leaves the partially downloaded file in place instead of deleting it.

1. Confirm that the partial file exists and note its current size.

   $ ls -lh largefile.zip
   -rw-r--r-- 1 user user 45M Dec  6 10:15 largefile.zip

   A size significantly smaller than the expected total indicates that the file is incomplete and safe to use as a resume starting point.

1. Resume the download from the last received byte using --continue-at - together with the same URL and filename.

   $ curl --continue-at - --remote-name "https://www.example.com/largefile.zip"
     % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                    Dload  Upload   Total   Spent    Left  Speed
     45  102M   47 48.0M    0     0  2390k      0  0:00:43  0:00:02  0:00:41 2600k
   ##### snipped #####
   100  102M  100  102M    0     0  2510k      0  0:00:41  0:00:41 --:--:-- 2890k

   If the server does not support HTTP Range or the file on the server has changed since the original download, the resumed content can be invalid; in that case, remove the partial file and repeat the download without --continue-at.

   The value --continue-at - automatically uses the current size of the local file as the starting offset, avoiding manual byte calculations.

1. Verify that the resumed download completed successfully by checking the final size and comparing a checksum against a trusted reference.

   $ ls -lh largefile.zip
   -rw-r--r-- 1 user user 102M Dec  6 10:20 largefile.zip
    
   $ sha256sum largefile.zip
   d4735e3a265e16eee03f59718b9b5d03d6e0f4f0a5c5e5d7ab3c9c5b1f3d5a9  largefile.zip

   Matching file size and checksum values confirm a correct resume; if verification fails, delete the file with rm largefile.zip and perform a fresh download from the beginning.