sgl-project/sglang · error · HTTPException

{e}

Error message

{e}

What it means

The /register_transfer_engine_info endpoint on the EngineInfoBootstrapServer wraps its registration logic in try/except; any exception while storing a rank's transfer-engine info (malformed payload, bad rank, internal state error) is logged and re-raised as HTTP 400 with the original message as detail.

Source

Thrown at python/sglang/srt/entrypoints/engine_info_bootstrap_server.py:72

                tp_rank = data["tp_rank"]
                info = data["transfer_engine_info"]
                session_id = info["session_id"]
                weights_info_dict = info["weights_info_dict"]

                with self.lock:
                    self.transfer_engine_info[tp_rank] = (
                        session_id,
                        weights_info_dict,
                    )

                logger.info(
                    f"Registered transfer engine info for tp_rank={tp_rank}, "
                    f"session_id={session_id}"
                )
                return PlainTextResponse("OK")
            except Exception as e:
                logger.error(f"Failed to register engine info: {e}")
                raise HTTPException(status_code=400, detail=str(e))

        @app.get("/get_transfer_engine_info")
        def get_transfer_engine_info(rank: int):
            if rank < 0:
                raise HTTPException(status_code=400, detail="Invalid rank parameter")

            with self.lock:
                info = self.transfer_engine_info.get(rank)

            if info is None:
                raise HTTPException(
                    status_code=404,
                    detail=f"No transfer engine info for rank {rank}",
                )

            return {"rank": rank, "remote_instance_transfer_engine_info": list(info)}

        config = uvicorn.Config(app, host=host, port=port, log_level="warning")

View on GitHub (pinned to 0132848349)

Solutions

1. Read the response detail — it echoes the underlying exception message which names the real problem.
2. Ensure the registering client and server run the same sglang version so the payload schema matches.
3. Don't call this endpoint manually; let the launcher's rank-0 workers perform registration.

import requests
r = requests.post(f"http://{host}:{port}/register_transfer_engine_info",
                  json=payload, timeout=10)
assert r.status_code == 200, r.text  # surface detail before server logs

resp = requests.post(url, json=payload)
if resp.status_code == 400:
    logger.error("register rejected: %s", resp.json().get("detail"))
    # fix payload per detail, then retry once

When it happens

Trigger: POSTing to /register_transfer_engine_info with a body that fails parsing/validation or triggers an internal error during registration; common when hand-testing the endpoint or when a rank's registration payload doesn't match the expected schema.

Common situations: Custom scripts poking the bootstrap server; version skew between the registering worker and the bootstrap server changing the payload schema; retrying a duplicate/rank-mismatched registration.

Related errors

• Invalid rank parameter
• No transfer engine info for rank {rank}
• Unknown serve backend {name!r}. Available values: {available
• Multiple distributions register serve backend {name!r}: {pro
• Failed to load serve backend {name!r} from {self._entry_poin

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/340cf83fc70505f9. Report an issue: GitHub.