[GH-ISSUE #1829] [Bug][TENT]: SGLang P/D intra-node communication fails due to wrong IPC address #521

New issue

Closed

opened 2026-10-29 21:15:50 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-29 21:15:50 +00:00

Owner

Originally created by @shuoerw on GitHub (Apr 7, 2026).
Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/1829

Summary

SGLang intra-node PD disaggregation does not work correctly in TENT.
The root cause appears to be that the transfer engine gives wrong address for the metadata buffer, so the Prefill node metadata is not delivered to the Decode node, which prevents the Decode node from starting properly.

Problem

Because the metadata is not propagated as expected, the Decode node gets stuck at:
decode.py#L988
github.com/sgl-project/sglang@f6e85676b5/python/sglang/srt/disaggregation/decode.py (L989)
As a result, the Decode node never begins processing.

The metadata is registered in Mooncake here:
github.com/kvcache-ai/Mooncake@6da25727f8/mooncake-transfer-engine/tent/src/transport/nvlink/nvlink_transport.cpp (L211)
The root cause was found and is discussed in https://github.com/kvcache-ai/Mooncake/pull/1622

Possible Fix

The same fix can be migrated to TENT:
https://github.com/kvcache-ai/Mooncake/pull/1622

@ishandhanani @TTThanos @stmatengss

Originally created by @shuoerw on GitHub (Apr 7, 2026). Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/1829 ## Summary SGLang intra-node PD disaggregation does not work correctly in TENT. The root cause appears to be that the transfer engine gives wrong address for the metadata buffer, so the Prefill node metadata is not delivered to the Decode node, which prevents the Decode node from starting properly. ## Problem Because the metadata is not propagated as expected, the Decode node gets stuck at: decode.py#L988 https://github.com/sgl-project/sglang/blob/f6e85676b578108207c7b7d4e3e8a65625e90f00/python/sglang/srt/disaggregation/decode.py#L989 As a result, the Decode node never begins processing. The metadata is registered in Mooncake here: https://github.com/kvcache-ai/Mooncake/blob/6da25727f8b4d605b2947a7a63d8274652f00f96/mooncake-transfer-engine/tent/src/transport/nvlink/nvlink_transport.cpp#L211 The root cause was found and is discussed in https://github.com/kvcache-ai/Mooncake/pull/1622 ## Possible Fix The same fix can be migrated to TENT: https://github.com/kvcache-ai/Mooncake/pull/1622 @ishandhanani @TTThanos @stmatengss

bsakaguchi

2026-10-29 21:15:50 +00:00