/v2/multipart error

Thiyaga Rajasekaran

we are using large upload for 1 gb file, the multipart is working fine however receiving an error when creating a node.

{"error": "Upstream service error (HTTP 500): {\"error\":\"Cannot get file info. File not assembled.\",\"request_id\":\"d9d24ced-746f-42bb-a989-c8c18e21e63e\"}","errorcode": 502,"timestamp": "2026-04-09T16:50:16.992955200Z"}

Upload Large File (200 OK )

• Send a POST request to api/v2/multipart containing:
  • file_name
  • file_size
  • mime_type
• Verify the request is successful.
• Split the file into blocks of the size specified by the response.
• For each block, send a POST request to api/v2/multipart/<upload_key>/<block_number> containing:
  • part_content: the file containing the block contents.
• Verify each request is successful.
• Send a POST request to api/v2/multipart/<upload_key>. Verify the request is successful.

Create Node (500 Error)

• Send a POST request to api/v2/nodes containing:
  • parent_id
  • type (144)
  • name
  • upload_key
• Verify a node was created correctly.

appuq

The approach appears correct, but you could try isolating the problem. If I were you, I would either do this with your OTCS team.

1. Make sure your REST API is talking to just a Single OTCS server. If, for example, your OTCS server is part of an LB mark all the other nodes down while you conduct this experiment.
2. Make sure your OTCS is writing to an External File System. In some organizations, the file content goes to OTAS even though that only happens when you actually create a node.
3. There is a directory that is used for files in transit, it is the key called the Upload directory, maybe it is E:\temp\Upload. When you upload the fisrt part, see if a crazy looking file lands there. Similarly, if you broke it into 4 parts, you should see the 4 parts there.You did not say you are closing the file handle, but I saw a call so I think you are closing it.
4. With this simplified setup, try your node creation if it still bombs. Open tickets with OT with thread logs and have them help you.
5. If it works more than likely the uploads are landing in different upload directories so making it into a common CIFS\SMB might be the way.
6. You can also turn sticky on your LB so your REST code will always send it to the same OTCS server.

PS : I got curious, so I am going to check the code . Perhaps the upload_key is a cacheid, and they may be assembling blobs in the database, so take the upload directory part may be irrelevant, in a way you won't see a physical file happening.