minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

deploy: D mr320/cg256/mf0.83 (KV pool 41.8M tok/worker); d1 HTTP 30001->30030 fix port collision with p1 scheduler loopback bind

Browse Source
This commit is contained in:
minke.yu
2026-09-28 20:00:11 +08:00
parent 735434fcde
commit e403cede0a
1 changed files with 451 additions and 451 deletions
Download Patch File Download Diff File Expand all files Collapse all files
deploy/b300-ds41/k8s/ds41-3p1d-rbg-cw.yaml
+7 -7
View File
@@ -1,5 +1,5 @@
# ds41-3p1d: DeepSeek-V4.1-Flash 3P1D PD 分离（cs32k + OTLP trace），RBG 版 # ds41-3p1d: DeepSeek-V4.1-Flash 3P1D PD 分离（cs32k + OTLP trace），RBG 版
# 拓扑: 3×prefill(tp2+ep2+cp2, cs32k, dspark, hicache L3) + 1×decode(tp2+ep2+dp2, cg512, dspark, # 拓扑: 3×prefill(tp2+ep2+cp2, cs32k, dspark, hicache L3) + 1×decode(tp2+ep2+dp2, cg256 mr320, dspark,
# load-balance-method=total_requests 按请求数分流) + mooncake-master(hicache L3 store) + router(mini-lb PD) # load-balance-method=total_requests 按请求数分流) + mooncake-master(hicache L3 store) + router(mini-lb PD)
# 全部钉 b300-01（单机 8 卡，hostNetwork + hostPID——mooncake NVLink-intra 跨 Pod CUDA IPC 必需， # 全部钉 b300-01（单机 8 卡，hostNetwork + hostPID——mooncake NVLink-intra 跨 Pod CUDA IPC 必需，
# 等同 docker compose 的 pid: host；参考 AGENTS.md「部署四要素」） # 等同 docker compose 的 pid: host；参考 AGENTS.md「部署四要素」）
@@ -154,16 +154,16 @@ spec:
--model-path /data/models/DeepSeek-V4.1-Flash \ --model-path /data/models/DeepSeek-V4.1-Flash \
--tp 2 --ep-size 2 --dp-size 2 \ --tp 2 --ep-size 2 --dp-size 2 \
--enable-dp-attention --enable-dp-lm-head \ --enable-dp-attention --enable-dp-lm-head \
--mem-fraction-static 0.80 \ --mem-fraction-static 0.83 \
--attention-backend dsv4 \ --attention-backend dsv4 \
--moe-runner-backend flashinfer_mxfp4 \ --moe-runner-backend flashinfer_mxfp4 \
--cuda-graph-max-bs-decode 512 \ --cuda-graph-max-bs-decode 256 \
--load-balance-method total_requests \ --load-balance-method total_requests \
--reasoning-parser auto --tool-call-parser auto \ --reasoning-parser auto --tool-call-parser auto \
--default-chat-template-kwargs '{"thinking": true, "reasoning_effort": "high"}' \ --default-chat-template-kwargs '{"thinking": true, "reasoning_effort": "high"}' \
--max-running-requests 256 \ --max-running-requests 320 \
--tokenizer-worker-num 8 \ --tokenizer-worker-num 8 \
--host 0.0.0.0 --port 30001 \ --host 0.0.0.0 --port 30030 \
--enable-cache-report --enable-metrics \ --enable-cache-report --enable-metrics \
--speculative-algorithm DSPARK --speculative-dspark-block-size 5 \ --speculative-algorithm DSPARK --speculative-dspark-block-size 5 \
--disaggregation-mode decode \ --disaggregation-mode decode \
@@ -194,7 +194,7 @@ spec:
value: Asia/Shanghai value: Asia/Shanghai
ports: ports:
- name: metrics - name: metrics
containerPort: 30001 containerPort: 30030
readinessProbe: readinessProbe:
httpGet: httpGet:
path: /health path: /health
@@ -399,7 +399,7 @@ spec:
--prefill http://192.168.1.243:30000 8998 \ --prefill http://192.168.1.243:30000 8998 \
--prefill http://192.168.1.243:30010 8999 \ --prefill http://192.168.1.243:30010 8999 \
--prefill http://192.168.1.243:30020 9000 \ --prefill http://192.168.1.243:30020 9000 \
--decode http://192.168.1.243:30001 --decode http://192.168.1.243:30030
ports: ports:
- name: http - name: http
containerPort: 50002 containerPort: 50002