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
+451 -451
View File
@@ -1,451 +1,451 @@
# 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」）
# 镜像已侧载 containerd: docker.io/ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix（IfNotPresent） # 镜像已侧载 containerd: docker.io/ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix（IfNotPresent）
# 参数来源: deploy/b300-ds41/dockerserve-3p1d-otlp-cs32k.yml（docker compose 版） # 参数来源: deploy/b300-ds41/dockerserve-3p1d-otlp-cs32k.yml（docker compose 版）
# 对外: 仅 ClusterIP Service ds41-3p1d:50002（router 用普通 Pod 网络，不占宿主机端口， # 对外: 仅 ClusterIP Service ds41-3p1d:50002（router 用普通 Pod 网络，不占宿主机端口，
# 天然不对宿主机/外网暴露）；new-api-gw 用集群 DNS http://ds41-3p1d:50002/v1 访问 # 天然不对宿主机/外网暴露）；new-api-gw 用集群 DNS http://ds41-3p1d:50002/v1 访问
# trace: OTLP → 127.0.0.1:4317（同节点 jaeger-trace cw，badger 6h 轮窗） # trace: OTLP → 127.0.0.1:4317（同节点 jaeger-trace cw，badger 6h 轮窗）
apiVersion: workloads.x-k8s.io/v1alpha2 apiVersion: workloads.x-k8s.io/v1alpha2
kind: RoleBasedGroup kind: RoleBasedGroup
metadata: metadata:
name: ds41-3p1d name: ds41-3p1d
labels: labels:
app: ds41-3p1d app: ds41-3p1d
sglang-model: ds-v41-flash sglang-model: ds-v41-flash
spec: spec:
roleTemplates: roleTemplates:
- name: prefill-base - name: prefill-base
template: template:
metadata: metadata:
labels: labels:
sglang-model: ds-v41-flash sglang-model: ds-v41-flash
sglang-role: prefill sglang-role: prefill
spec: spec:
hostNetwork: true hostNetwork: true
hostPID: true hostPID: true
dnsPolicy: ClusterFirstWithHostNet dnsPolicy: ClusterFirstWithHostNet
nodeSelector: nodeSelector:
kubernetes.io/hostname: b300-01 kubernetes.io/hostname: b300-01
containers: containers:
- name: sglang - name: sglang
image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix
imagePullPolicy: IfNotPresent imagePullPolicy: IfNotPresent
command: command:
- /bin/bash - /bin/bash
- -c - -c
- | - |
exec sglang serve \ exec sglang serve \
--trust-remote-code \ --trust-remote-code \
--model-path /data/models/DeepSeek-V4.1-Flash \ --model-path /data/models/DeepSeek-V4.1-Flash \
--tp 2 --ep-size 2 \ --tp 2 --ep-size 2 \
--enable-prefill-cp --cp-strategy interleave \ --enable-prefill-cp --cp-strategy interleave \
--chunked-prefill-size 32768 --max-prefill-tokens 32768 \ --chunked-prefill-size 32768 --max-prefill-tokens 32768 \
--mem-fraction-static 0.75 \ --mem-fraction-static 0.75 \
--attention-backend dsv4 \ --attention-backend dsv4 \
--moe-runner-backend flashinfer_mxfp4 \ --moe-runner-backend flashinfer_mxfp4 \
--cuda-graph-max-bs-decode 32 \ --cuda-graph-max-bs-decode 32 \
--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 256 \
--tokenizer-worker-num 8 \ --tokenizer-worker-num 8 \
--host 0.0.0.0 --port "$SGLANG_PORT" \ --host 0.0.0.0 --port "$SGLANG_PORT" \
--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 \
--enable-hierarchical-cache \ --enable-hierarchical-cache \
--hicache-ratio 2 \ --hicache-ratio 2 \
--hicache-mem-layout page_first_direct \ --hicache-mem-layout page_first_direct \
--hicache-io-backend direct \ --hicache-io-backend direct \
--hicache-write-policy write_through \ --hicache-write-policy write_through \
--hicache-storage-backend mooncake \ --hicache-storage-backend mooncake \
--hicache-storage-prefetch-policy wait_complete \ --hicache-storage-prefetch-policy wait_complete \
--hicache-size 0 \ --hicache-size 0 \
--disaggregation-mode prefill \ --disaggregation-mode prefill \
--disaggregation-transfer-backend mooncake \ --disaggregation-transfer-backend mooncake \
--disaggregation-bootstrap-port "$SGLANG_BOOTSTRAP_PORT" \ --disaggregation-bootstrap-port "$SGLANG_BOOTSTRAP_PORT" \
--enable-trace \ --enable-trace \
--trace-modules request,mooncake \ --trace-modules request,mooncake \
--otlp-traces-endpoint 127.0.0.1:4317 \ --otlp-traces-endpoint 127.0.0.1:4317 \
--otlp-service-name "$OTLP_NAME" --otlp-service-name "$OTLP_NAME"
env: env:
- name: SGLANG_TORCH_PROFILER_DIR - name: SGLANG_TORCH_PROFILER_DIR
value: /data/ymk/prof/torch value: /data/ymk/prof/torch
- name: SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE - name: SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE
value: "1" value: "1"
- name: SGLANG_RAGGED_VERIFY_MODE - name: SGLANG_RAGGED_VERIFY_MODE
value: static value: static
- name: MC_INTRANODE_NVLINK - name: MC_INTRANODE_NVLINK
value: "true" value: "true"
- name: MC_INTRA_NVLINK - name: MC_INTRA_NVLINK
value: "true" value: "true"
- name: SGLANG_MOONCAKE_SEND_AUX_TCP - name: SGLANG_MOONCAKE_SEND_AUX_TCP
value: "1" value: "1"
- name: SGLANG_HICACHE_MOONCAKE_CONFIG_PATH - name: SGLANG_HICACHE_MOONCAKE_CONFIG_PATH
value: /data/ymk/ds41/mooncake-store.json value: /data/ymk/ds41/mooncake-store.json
- name: SGLANG_DISAGGREGATION_QUEUE_SIZE - name: SGLANG_DISAGGREGATION_QUEUE_SIZE
value: "16" value: "16"
- name: SGLANG_DISAGGREGATION_THREAD_POOL_SIZE - name: SGLANG_DISAGGREGATION_THREAD_POOL_SIZE
value: "32" value: "32"
- name: TZ - name: TZ
value: Asia/Shanghai value: Asia/Shanghai
readinessProbe: readinessProbe:
httpGet: httpGet:
path: /health path: /health
port: metrics port: metrics
initialDelaySeconds: 60 initialDelaySeconds: 60
periodSeconds: 10 periodSeconds: 10
timeoutSeconds: 5 timeoutSeconds: 5
failureThreshold: 90 failureThreshold: 90
resources: resources:
requests: requests:
cpu: "32" cpu: "32"
memory: 220Gi memory: 220Gi
nvidia.com/gpu: "2" nvidia.com/gpu: "2"
limits: limits:
cpu: "96" cpu: "96"
memory: 950Gi memory: 950Gi
nvidia.com/gpu: "2" nvidia.com/gpu: "2"
securityContext: securityContext:
privileged: true privileged: true
volumeMounts: volumeMounts:
- name: data - name: data
mountPath: /data mountPath: /data
- name: sglang-cache - name: sglang-cache
mountPath: /root/.cache/sglang mountPath: /root/.cache/sglang
- name: dshm - name: dshm
mountPath: /dev/shm mountPath: /dev/shm
volumes: volumes:
- name: data - name: data
hostPath: hostPath:
path: /data path: /data
type: Directory type: Directory
- name: sglang-cache - name: sglang-cache
hostPath: hostPath:
path: /data/ymk/cache/sglang path: /data/ymk/cache/sglang
type: Directory type: Directory
- name: dshm - name: dshm
emptyDir: emptyDir:
medium: Memory medium: Memory
sizeLimit: 32Gi sizeLimit: 32Gi
- name: decode-base - name: decode-base
template: template:
metadata: metadata:
labels: labels:
sglang-model: ds-v41-flash sglang-model: ds-v41-flash
sglang-role: decode sglang-role: decode
spec: spec:
hostNetwork: true hostNetwork: true
hostPID: true hostPID: true
dnsPolicy: ClusterFirstWithHostNet dnsPolicy: ClusterFirstWithHostNet
nodeSelector: nodeSelector:
kubernetes.io/hostname: b300-01 kubernetes.io/hostname: b300-01
containers: containers:
- name: sglang - name: sglang
image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix
imagePullPolicy: IfNotPresent imagePullPolicy: IfNotPresent
command: command:
- /bin/bash - /bin/bash
- -c - -c
- | - |
exec sglang serve \ exec sglang serve \
--trust-remote-code \ --trust-remote-code \
--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 \
--disaggregation-transfer-backend mooncake \ --disaggregation-transfer-backend mooncake \
--disaggregation-bootstrap-port 9001 \ --disaggregation-bootstrap-port 9001 \
--enable-trace \ --enable-trace \
--trace-modules request,mooncake \ --trace-modules request,mooncake \
--otlp-traces-endpoint 127.0.0.1:4317 \ --otlp-traces-endpoint 127.0.0.1:4317 \
--otlp-service-name ds41-3p1d-d1 --otlp-service-name ds41-3p1d-d1
env: env:
- name: SGLANG_TORCH_PROFILER_DIR - name: SGLANG_TORCH_PROFILER_DIR
value: /data/ymk/prof/torch value: /data/ymk/prof/torch
- name: SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE - name: SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE
value: "1" value: "1"
- name: SGLANG_RAGGED_VERIFY_MODE - name: SGLANG_RAGGED_VERIFY_MODE
value: static value: static
- name: MC_INTRANODE_NVLINK - name: MC_INTRANODE_NVLINK
value: "true" value: "true"
- name: MC_INTRA_NVLINK - name: MC_INTRA_NVLINK
value: "true" value: "true"
- name: SGLANG_MOONCAKE_SEND_AUX_TCP - name: SGLANG_MOONCAKE_SEND_AUX_TCP
value: "1" value: "1"
- name: SGLANG_DISAGGREGATION_QUEUE_SIZE - name: SGLANG_DISAGGREGATION_QUEUE_SIZE
value: "16" value: "16"
- name: SGLANG_DISAGGREGATION_THREAD_POOL_SIZE - name: SGLANG_DISAGGREGATION_THREAD_POOL_SIZE
value: "32" value: "32"
- name: TZ - name: TZ
value: Asia/Shanghai value: Asia/Shanghai
ports: ports:
- name: metrics - name: metrics
containerPort: 30001 containerPort: 30030
readinessProbe: readinessProbe:
httpGet: httpGet:
path: /health path: /health
port: metrics port: metrics
initialDelaySeconds: 60 initialDelaySeconds: 60
periodSeconds: 10 periodSeconds: 10
timeoutSeconds: 5 timeoutSeconds: 5
failureThreshold: 90 failureThreshold: 90
resources: resources:
requests: requests:
cpu: "32" cpu: "32"
memory: 100Gi memory: 100Gi
nvidia.com/gpu: "2" nvidia.com/gpu: "2"
limits: limits:
cpu: "96" cpu: "96"
memory: 600Gi memory: 600Gi
nvidia.com/gpu: "2" nvidia.com/gpu: "2"
securityContext: securityContext:
privileged: true privileged: true
volumeMounts: volumeMounts:
- name: data - name: data
mountPath: /data mountPath: /data
- name: sglang-cache - name: sglang-cache
mountPath: /root/.cache/sglang mountPath: /root/.cache/sglang
- name: dshm - name: dshm
mountPath: /dev/shm mountPath: /dev/shm
volumes: volumes:
- name: data - name: data
hostPath: hostPath:
path: /data path: /data
type: Directory type: Directory
- name: sglang-cache - name: sglang-cache
hostPath: hostPath:
path: /data/ymk/cache/sglang path: /data/ymk/cache/sglang
type: Directory type: Directory
- name: dshm - name: dshm
emptyDir: emptyDir:
medium: Memory medium: Memory
sizeLimit: 32Gi sizeLimit: 32Gi
roles: roles:
- name: mooncake-master - name: mooncake-master
replicas: 1 replicas: 1
standalonePattern: standalonePattern:
template: template:
metadata: metadata:
labels: labels:
app: ds41-3p1d-mooncake-master app: ds41-3p1d-mooncake-master
spec: spec:
hostNetwork: true hostNetwork: true
dnsPolicy: ClusterFirstWithHostNet dnsPolicy: ClusterFirstWithHostNet
nodeSelector: nodeSelector:
kubernetes.io/hostname: b300-01 kubernetes.io/hostname: b300-01
containers: containers:
- name: mooncake-master - name: mooncake-master
image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix
imagePullPolicy: IfNotPresent imagePullPolicy: IfNotPresent
command: command:
- /bin/bash - /bin/bash
- -c - -c
- | - |
exec /opt/sglang/bin/mooncake_master \ exec /opt/sglang/bin/mooncake_master \
--port 50051 \ --port 50051 \
--enable_http_metadata_server=true \ --enable_http_metadata_server=true \
--http_metadata_server_port=18080 \ --http_metadata_server_port=18080 \
--eviction_high_watermark_ratio=0.90 --eviction_high_watermark_ratio=0.90
readinessProbe: readinessProbe:
tcpSocket: tcpSocket:
port: 50051 port: 50051
initialDelaySeconds: 5 initialDelaySeconds: 5
periodSeconds: 5 periodSeconds: 5
failureThreshold: 60 failureThreshold: 60
resources: resources:
requests: requests:
cpu: "1" cpu: "1"
memory: 8Gi memory: 8Gi
limits: limits:
cpu: "4" cpu: "4"
memory: 64Gi memory: 64Gi
- name: p1 - name: p1
replicas: 1 replicas: 1
dependencies: dependencies:
- mooncake-master - mooncake-master
standalonePattern: standalonePattern:
templateRef: templateRef:
name: prefill-base name: prefill-base
patch: patch:
metadata: metadata:
labels: labels:
app: ds41-3p1d-p1 app: ds41-3p1d-p1
spec: spec:
containers: containers:
- name: sglang - name: sglang
env: env:
- name: CUDA_VISIBLE_DEVICES - name: CUDA_VISIBLE_DEVICES
value: "0,1" value: "0,1"
- name: SGLANG_PORT - name: SGLANG_PORT
value: "30000" value: "30000"
- name: SGLANG_BOOTSTRAP_PORT - name: SGLANG_BOOTSTRAP_PORT
value: "8998" value: "8998"
- name: OTLP_NAME - name: OTLP_NAME
value: ds41-3p1d-p1 value: ds41-3p1d-p1
ports: ports:
- name: metrics - name: metrics
containerPort: 30000 containerPort: 30000
- name: p2 - name: p2
replicas: 1 replicas: 1
dependencies: dependencies:
- mooncake-master - mooncake-master
standalonePattern: standalonePattern:
templateRef: templateRef:
name: prefill-base name: prefill-base
patch: patch:
metadata: metadata:
labels: labels:
app: ds41-3p1d-p2 app: ds41-3p1d-p2
spec: spec:
containers: containers:
- name: sglang - name: sglang
env: env:
- name: CUDA_VISIBLE_DEVICES - name: CUDA_VISIBLE_DEVICES
value: "2,3" value: "2,3"
- name: SGLANG_PORT - name: SGLANG_PORT
value: "30010" value: "30010"
- name: SGLANG_BOOTSTRAP_PORT - name: SGLANG_BOOTSTRAP_PORT
value: "8999" value: "8999"
- name: OTLP_NAME - name: OTLP_NAME
value: ds41-3p1d-p2 value: ds41-3p1d-p2
ports: ports:
- name: metrics - name: metrics
containerPort: 30010 containerPort: 30010
- name: p3 - name: p3
replicas: 1 replicas: 1
dependencies: dependencies:
- mooncake-master - mooncake-master
standalonePattern: standalonePattern:
templateRef: templateRef:
name: prefill-base name: prefill-base
patch: patch:
metadata: metadata:
labels: labels:
app: ds41-3p1d-p3 app: ds41-3p1d-p3
spec: spec:
containers: containers:
- name: sglang - name: sglang
env: env:
- name: CUDA_VISIBLE_DEVICES - name: CUDA_VISIBLE_DEVICES
value: "4,5" value: "4,5"
- name: SGLANG_PORT - name: SGLANG_PORT
value: "30020" value: "30020"
- name: SGLANG_BOOTSTRAP_PORT - name: SGLANG_BOOTSTRAP_PORT
value: "9000" value: "9000"
- name: OTLP_NAME - name: OTLP_NAME
value: ds41-3p1d-p3 value: ds41-3p1d-p3
ports: ports:
- name: metrics - name: metrics
containerPort: 30020 containerPort: 30020
- name: d1 - name: d1
replicas: 1 replicas: 1
dependencies: dependencies:
- mooncake-master - mooncake-master
standalonePattern: standalonePattern:
templateRef: templateRef:
name: decode-base name: decode-base
patch: patch:
metadata: metadata:
labels: labels:
app: ds41-3p1d-d1 app: ds41-3p1d-d1
spec: spec:
containers: containers:
- name: sglang - name: sglang
env: env:
- name: CUDA_VISIBLE_DEVICES - name: CUDA_VISIBLE_DEVICES
value: "6,7" value: "6,7"
- name: router - name: router
replicas: 1 replicas: 1
dependencies: dependencies:
- p1 - p1
- p2 - p2
- p3 - p3
- d1 - d1
standalonePattern: standalonePattern:
template: template:
metadata: metadata:
labels: labels:
app: ds41-3p1d-router app: ds41-3p1d-router
spec: spec:
# 入口不走 hostNetwork：普通 Pod 网络 + ClusterIP，天然不对宿主机/外网暴露。 # 入口不走 hostNetwork：普通 Pod 网络 + ClusterIP，天然不对宿主机/外网暴露。
# P/D 地址用节点内网 IP（hostNetwork 监听在 192.168.1.243 上，Calico 可达）。 # P/D 地址用节点内网 IP（hostNetwork 监听在 192.168.1.243 上，Calico 可达）。
nodeSelector: nodeSelector:
kubernetes.io/hostname: b300-01 kubernetes.io/hostname: b300-01
containers: containers:
- name: router - name: router
image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix image: ymkymx/sglang:dsv41-pd-visioncp-db7d2cb7d-fix
imagePullPolicy: IfNotPresent imagePullPolicy: IfNotPresent
command: command:
- /bin/bash - /bin/bash
- -c - -c
- | - |
exec sglang-router launch \ exec sglang-router launch \
--mini-lb --pd-disaggregation \ --mini-lb --pd-disaggregation \
--host 0.0.0.0 --port 50002 \ --host 0.0.0.0 --port 50002 \
--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
readinessProbe: readinessProbe:
httpGet: httpGet:
path: /health path: /health
port: http port: http
initialDelaySeconds: 5 initialDelaySeconds: 5
periodSeconds: 10 periodSeconds: 10
timeoutSeconds: 5 timeoutSeconds: 5
failureThreshold: 30 failureThreshold: 30
resources: resources:
requests: requests:
cpu: "1" cpu: "1"
memory: 4Gi memory: 4Gi
limits: limits:
cpu: "8" cpu: "8"
memory: 16Gi memory: 16Gi
--- ---
apiVersion: v1 apiVersion: v1
kind: Service kind: Service
metadata: metadata:
name: ds41-3p1d name: ds41-3p1d
labels: labels:
app: ds41-3p1d app: ds41-3p1d
spec: spec:
type: ClusterIP type: ClusterIP
selector: selector:
rbg.workloads.x-k8s.io/group-name: ds41-3p1d rbg.workloads.x-k8s.io/group-name: ds41-3p1d
rbg.workloads.x-k8s.io/role-name: router rbg.workloads.x-k8s.io/role-name: router
ports: ports:
- name: http - name: http
port: 50002 port: 50002
targetPort: 50002 targetPort: 50002
--- ---
apiVersion: monitoring.coreos.com/v1 apiVersion: monitoring.coreos.com/v1
kind: PodMonitor kind: PodMonitor
metadata: metadata:
name: ds41-3p1d name: ds41-3p1d
labels: labels:
release: kube-prometheus-stack release: kube-prometheus-stack
spec: spec:
selector: selector:
matchLabels: matchLabels:
sglang-model: ds-v41-flash sglang-model: ds-v41-flash
podMetricsEndpoints: podMetricsEndpoints:
- interval: 15s - interval: 15s
path: /metrics path: /metrics
port: metrics port: metrics