A frozen node is twice as expensive as an absent one, and marks its peer's completed work as failed
維護者通常 5 天內回覆
@Noah-Tervalon-Nvidia 已經在處理了。
開始於 2026年9月11日。
評估
這個 Issue 還沒有評估資料。
描述
Two observations from the same failure mode, measured on a two-node cluster (Ubuntu + RTX 3060, macOS, v0.1.1-463). Both matter for a cluster that includes a laptop, since a laptop that sleeps mid-request is exactly this shape: the socket is accepted, nothing answers.
Setup
24 concurrent chat/completions, max_tokens: 200, google/gemma-4-12b resident on both nodes.
| Condition, same load | Wall clock |
|---|---|
| Both nodes healthy | 37.8 s |
| Single node, bypassing the router entirely | 72.8 s |
Peer frozen (SIGSTOP on its nvpair processes) |
155.6 s |
Every request returned 200 in all three. Reliability is not the problem — the cost is.
A frozen peer is 2.1× worse than not having it at all, and 4.1× worse than a healthy cluster. When I instead killed the peer's processes outright, the same load finished in 28.3 s with no failures: a closed socket gives an immediate RST and the proxy moves on in milliseconds. It is specifically the accepted-but-silent socket that hurts.
For context on the eviction path: after the peer was killed, discovery had dropped it from the node list within 75 s — I sampled too coarsely to pin it more precisely. During that window each request pays its own connection timeout.
The second half: false failures
When the frozen peer resumed, its job history marked several of the other node's workloads as
Failed — node serveri stopped reporting this workload
Those workloads had all completed successfully; the requesting node returned 200 for all 24. The peer's view had simply gone stale while it was stopped, and on resume it reaped them as failed. The timestamps match the resume to the second.
That leaves the job history — the only place a user sees what the cluster did — showing failures that did not happen, precisely in the situation where something unusual was going on and they would be looking at it.
Not filed as a defect
Neither is a crash and both may be deliberate. But if a laptop is a first-class node, asleep with the socket accepted and nothing answering is its most likely state, not an edge case. Two directions that would help, if either is welcome:
- a connection-level timeout distinctly shorter than the request timeout, so an unresponsive candidate is abandoned quickly while a slow-but-live generation is not
- reaping a peer's workloads as
unknownrather thanfailedwhen the gap was on our side
Happy to open a PR for either if you have a preferred shape.
- 主要語言
- Go
- 星號
- 1.6k
- 分支
- 266
- 平均合併
- 3 天 6 小時
- 30 天內合併 PR
- 22
環境準備
- 沒有 Dockerfile 或 Docker Compose 檔案
- 有 Pull Request 範本
- 閱讀貢獻指南
從這裡開始
- 先讀完整個 Issue，再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫，在一個分支上完成修改。
- 送出 Pull Request，並在描述裡引用這個 Issue 編號。
NVIDIA/Personal-AI-Router 的其他 Issue
-
難度 5/5 一週以上 新手友好度 20/100
NVIDIA/Personal-AI-Router#165 ·
維護者通常 5 天內回覆
-
bug
難度 4/5 3-5 天 新手友好度 46/100
NVIDIA/Personal-AI-Router#164 ·
維護者通常 5 天內回覆
-
enhancement
難度 4/5 一週以上 新手友好度 12/100
NVIDIA/Personal-AI-Router#162 ·
維護者通常 5 天內回覆
-
enhancement
難度 4/5 3-5 天 新手友好度 25/100
NVIDIA/Personal-AI-Router#154 · 1 則留言 ·
維護者通常 5 天內回覆
-
[Feature]: [Ollama] Alert the user about model pull failures可能已有人在做 @ckelseynv 於 4 天前認領。 未關閉enhancement
難度 3/5 1-2 天 新手友好度 56/100
NVIDIA/Personal-AI-Router#152 · 1 則留言 · 已指派 1 人 ·
維護者通常 5 天內回覆
查看 NVIDIA/Personal-AI-Router 的全部 Issue
相似的 Issue
-
difficulty: easy documentation
難度 2/5 1-3 小時 新手友好度 70/100
維護者通常 1 天內回覆
-
難度 1/5 1 小時以內 新手友好度 85/100
-
難度 2/5 1-3 小時 新手友好度 74/100
scaleway/terraform-provider-scaleway#4485 ·
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 65/100
維護者通常 2 天內回覆
-
難度 2/5 1-3 小時 新手友好度 75/100