LWS 기반 PD 배포
LWS 기반 PD 배포 (LWS Based PD Deploy)
이 페이지는 Kubernetes의 LWS(LeaderWorkerSet)를 사용해 SGLang의 PD(Prefill/Decode) 분리 배포를 구성하는 방법을 설명해요. Prefill 서버와 Decode 서버를 각각의 LeaderWorkerSet으로 배포하고, 로드 밸런서(minilb)를 두어 사용자가 모델 API에 직접 접근할 수 있게 해요.
출처: 문서
본문
0. 사전 조건 (Prerequisites)
- k8s >= 1.26
- k8s에 lws 설치
1. 이미지 준비 (Image Preparation)
lmsysorg/sglang:deepep
2. 배포 매니페스트 파일 (Deployment Manifest Files)
참고: 곧 모든 배포 파일을 Helm Chart 형식으로 패키징할 예정이에요. 관심 있는 커뮤니티 구성원은 연락해 기여할 수 있어요.
Prefill
Prefill 매니페스트 파일 prefill.yaml
참고: NodeSelector 섹션, 모델 위치 섹션, taint toleration 섹션은 실제 배포 환경에 따라 조정할 수 있어요.
apiVersion: leaderworkerset.x-k8s.io/v1
kind: LeaderWorkerSet
metadata:
name: deepseekr10528-prefill-main
spec:
leaderWorkerTemplate:
leaderTemplate:
metadata:
labels:
role: leader
spec:
containers:
- command:
- python3
- -m
- sglang.launch_server
- --port
- "30000"
- --host
- "0.0.0.0"
- --model-path
- /work/models
- --disaggregation-ib-device
# should modify according your rdma env
- mlx5_bond_0,mlx5_bond_1,mlx5_bond_2,mlx5_bond_3
- --chunked-prefill-size
- "524288"
- --max-prefill-tokens
- "32768"
- --page-size
- "64"
# - --init-expert-location
# - /home/aiges/tuned/attachment_ep_statistics/prefill_in1024.json
- --ep-dispatch-algorithm
- dynamic
- --eplb-algorithm
- deepseek
# - --deepep-config
# - /home/aiges/tuned/tuned_8sms.json
- --enable-dp-lm-head
- --enable-dp-attention
- --dp-size
- "16"
- --disable-radix-cache
- --moe-a2a-backend
- deepep
- --disaggregation-mode
- prefill
- --mem-fraction-static
- "0.7"
- --context-length
- "32768"
- --tp
- "16"
- --dist-init-addr
- $(LWS_LEADER_ADDRESS):20102
- --nnodes
- $(LWS_GROUP_SIZE)
- --node-rank
- $(LWS_WORKER_INDEX)
- --trust-remote-code
- --ep-num-redundant-experts
- "32"
- --moe-dense-tp-size
- "1"
- --max-running-requests
- "1024"
env:
- name: NVSHMEM_IB_GID_INDEX
value: "3"
- name: NVSHMEM_ENABLE_NIC_PE_MAPPING
value: "1"
- name: SGLANG_SET_CPU_AFFINITY
value: "true"
- name: SGLANG_ENABLE_JIT_DEEPGEMM
value: "1"
- name: NCCL_IB_QPS_PER_CONNECTION
value: "8"
- name: NCCL_IB_SPLIT_DATA_ON_QPS
value: "1"
- name: NCCL_NET_PLUGIN
value: none
- name: NCCL_IB_TC
value: "136"
- name: NCCL_MIN_NCHANNELS
value: "4"
- name: MC_TE_METRIC
value: "false"
- name: NCCL_IB_SL
value: "5"
- name: NCCL_IB_HCA
value: ^=mlx5_0,mlx5_5,mlx5_6
- name: LWS_WORKER_INDEX
valueFrom:
fieldRef:
fieldPath: metadata.labels['leaderworkerset.sigs.k8s.io/worker-index']
image: lmsysorg/sglang:deepep
name: sglang-leader
ports:
- containerPort: 30000
protocol: TCP
readinessProbe:
periodSeconds: 30
tcpSocket:
port: 30000
resources:
limits:
nvidia.com/gpu: "8"
securityContext:
capabilities:
add:
- IPC_LOCK
privileged: true
volumeMounts:
- mountPath: /dev/shm
name: dshm
- mountPath: /work/models
name: model
- mountPath: /dev/infiniband
name: ib
- mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs
name: cf
- mountPath: /root/.cache
name: sgl-cache
dnsPolicy: ClusterFirstWithHostNet
hostIPC: true
hostNetwork: true
nodeSelector:
pd: "yes"
tolerations:
- key: pd
operator: Exists
- key: node-role
operator: Exists
volumes:
- emptyDir:
medium: Memory
name: dshm
- hostPath:
# modify according to you deployment env
path: /data1/maas_hosted_models/models/DeepSeek-R1-0528/deepseek_r1_0528
name: model
- hostPath:
path: /dev/infiniband
name: ib
- hostPath:
# modify according to you deployment env
path: /data1/maas_hosted_models/models/fused_moe_triton/configs
name: cf
- hostPath:
# modify according to you deployment env
path: /data1/sgl_cache
type: DirectoryOrCreate
name: sgl-cache
restartPolicy: RecreateGroupOnPodRestart
size: 2
workerTemplate:
metadata: {}
spec:
containers:
- command:
- python3
- -m
- sglang.launch_server
- --model-path
- /work/models
- --disaggregation-ib-device
- mlx5_bond_0,mlx5_bond_1,mlx5_bond_2,mlx5_bond_3
- --chunked-prefill-size
- "524288"
- --max-prefill-tokens
- "32768"
- --page-size
- "64"
- --ep-dispatch-algorithm
- dynamic
- --eplb-algorithm
- deepseek
- --enable-dp-lm-head
- --enable-dp-attention
- --dp-size
- "16"
- --disable-radix-cache
- --moe-a2a-backend
- deepep
- --disaggregation-mode
- prefill
- --mem-fraction-static
- "0.7"
- --context-length
- "32768"
- --tp
- "16"
- --dist-init-addr
- $(LWS_LEADER_ADDRESS):20102
- --nnodes
- $(LWS_GROUP_SIZE)
- --node-rank
- $(LWS_WORKER_INDEX)
- --trust-remote-code
- --ep-num-redundant-experts
- "32"
- --moe-dense-tp-size
- "1"
- --max-running-requests
- "1024"
env:
- name: SGLANG_SET_CPU_AFFINITY
value: "true"
- name: SGLANG_HACK_DEEPEP_NUM_SMS
value: "8"
- name: SGLANG_HACK_DEEPEP_NEW_MODE
value: "0"
- name: NCCL_IB_HCA
value: ^=mlx5_0,mlx5_5,mlx5_6
- name: NVSHMEM_IB_TRAFFIC_CLASS
value: "16"
- name: NVSHMEM_IB_GID_INDEX
value: "3"
- name: NVSHMEM_ENABLE_NIC_PE_MAPPING
value: "1"
- name: CUDA_LAUNCH_BLOCKING
value: "0"
- name: SGLANG_MOONCAKE_TRANS_THREAD
value: "8"
- name: SGLANG_ENABLE_JIT_DEEPGEMM
value: "1"
- name: SGLANG_CHUNKED_PREFIX_CACHE_THRESHOLD
value: "0"
- name: NCCL_IB_QPS_PER_CONNECTION
value: "8"
- name: NCCL_IB_SPLIT_DATA_ON_QPS
value: "1"
- name: NCCL_NET_PLUGIN
value: none
- name: NCCL_IB_TC
value: "136"
- name: NCCL_MIN_NCHANNELS
value: "4"
- name: MC_TE_METRIC
value: "true"
- name: NCCL_IB_SL
value: "5"
- name: LWS_WORKER_INDEX
valueFrom:
fieldRef:
fieldPath: metadata.labels['leaderworkerset.sigs.k8s.io/worker-index']
image: lmsysorg/sglang:deepep
name: sglang-worker
ports:
- containerPort: 30001
protocol: TCP
resources:
limits:
nvidia.com/gpu: "8"
securityContext:
capabilities:
add:
- IPC_LOCK
privileged: true
volumeMounts:
- mountPath: /root/.cache
name: sgl-cache
- mountPath: /dev/shm
name: dshm
- mountPath: /work/models
name: model
- mountPath: /dev/infiniband
name: ib
- mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs
name: cf
dnsPolicy: ClusterFirstWithHostNet
hostIPC: true
hostNetwork: true
nodeSelector:
pd: "yes"
tolerations:
- key: pd
operator: Exists
- key: node-role
operator: Exists
volumes:
- emptyDir:
medium: Memory
name: dshm
- hostPath:
path: /dev/infiniband
name: ib
- hostPath:
path: /data1/maas_hosted_models/models/DeepSeek-R1-0528/deepseek_r1_0528
name: model
- hostPath:
path: /data1/maas_hosted_models/models/fused_moe_triton/configs
name: cf
- hostPath:
path: /data1/sgl_cache
type: DirectoryOrCreate
name: sgl-cache
Decode
Decode 노드 배포 매니페스트 파일 decode.yaml
참고: NodeSelector 섹션, 모델 위치 섹션, taint toleration 섹션은 실제 배포 환경에 따라 조정할 수 있어요.
apiVersion: leaderworkerset.x-k8s.io/v1
kind: LeaderWorkerSet
metadata:
name: deepseekr10528-decode-main
spec:
leaderWorkerTemplate:
leaderTemplate:
metadata:
labels:
role: leader
spec:
containers:
- command:
- python3
- -m
- sglang.launch_server
- --port
- "30000"
- --host
- "0.0.0.0"
- --model-path
- /work/models
- --chunked-prefill-size
- "262144"
- --page-size
- "64"
- --enable-dp-attention
- --enable-dp-lm-head
- --dp-size
- "16"
- --moe-a2a-backend
- deepep
- --disaggregation-mode
- decode
- --mem-fraction-static
- "0.849"
- --context-length
- "32768"
- --disaggregation-ib-device
- "mlx5_bond_0,mlx5_bond_1,mlx5_bond_2,mlx5_bond_3"
- --cuda-graph-max-bs-decode
- "64"
- --max-running-requests
- "2048"
- --tp-size
- "16" # Size of Tensor Parallelism
- --dist-init-addr
- $(LWS_LEADER_ADDRESS):20102
- --nnodes
- $(LWS_GROUP_SIZE)
- --node-rank
- $(LWS_WORKER_INDEX)
- --trust-remote-code
- --ep-num-redundant-experts
- "32"
- --moe-dense-tp-size
- "1"
env:
- name: CUDA_LAUNCH_BLOCKING
value: "0"
- name: NVSHMEM_IB_GID_INDEX
value: "3"
- name: NVSHMEM_ENABLE_NIC_PE_MAPPING
value: "1"
- name: NCCL_IB_QPS_PER_CONNECTION
value: "8"
- name: NCCL_IB_SPLIT_DATA_ON_QPS
value: "1"
- name: NCCL_NET_PLUGIN
value: "none"
- name: NCCL_IB_TC
value: "136"
- name: NCCL_MIN_NCHANNELS
value: "4"
- name: NCCL_IB_SL
value: "5"
- name: MC_TE_METRIC
value: "true"
- name: SGLANG_MOONCAKE_TRANS_THREAD
value: "16"
- name: SGLANG_ENABLE_JIT_DEEPGEMM
value: "1"
- name: NCCL_IB_HCA
value: ^=mlx5_0,mlx5_5,mlx5_6
- name: LWS_WORKER_INDEX
valueFrom:
fieldRef:
fieldPath: metadata.labels['leaderworkerset.sigs.k8s.io/worker-index']
image: lmsysorg/sglang:deepep
name: sglang-leader
ports:
- containerPort: 30000
protocol: TCP
readinessProbe:
periodSeconds: 30
tcpSocket:
port: 30000
resources:
limits:
nvidia.com/gpu: "8"
securityContext:
capabilities:
add:
- IPC_LOCK
privileged: true
volumeMounts:
- mountPath: /root/.cache
name: sgl-cache
- mountPath: /dev/shm
name: dshm
- mountPath: /work/models
name: model
- mountPath: /dev/infiniband
name: ib
- mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs
name: cf
dnsPolicy: ClusterFirstWithHostNet
hostIPC: true
hostNetwork: true
nodeSelector:
pd: "yes"
tolerations:
- key: pd
operator: Exists
- key: node-role
operator: Exists
volumes:
- hostPath:
path: /data1/sgl_cache1
type: DirectoryOrCreate
name: sgl-cache
- emptyDir:
medium: Memory
name: dshm
- hostPath:
path: /data1/maas_hosted_models/models/DeepSeek-R1-0528/deepseek_r1_0528
name: model
- hostPath:
path: /dev/infiniband
name: ib
- hostPath:
path: /data1/maas_hosted_models/models/fused_moe_triton/configs
name: cf
restartPolicy: RecreateGroupOnPodRestart
size: 2
workerTemplate:
metadata: {}
spec:
containers:
- command:
- python3
- -m
- sglang.launch_server
- --model-path
- /work/models
- --chunked-prefill-size
- "262144"
- --page-size
- "64"
- --enable-dp-attention
- --enable-dp-lm-head
- --dp-size
- "16"
- --moe-a2a-backend
- deepep
- --disaggregation-mode
- decode
- --mem-fraction-static
- "0.849"
- --context-length
- "32768"
- --disaggregation-ib-device
- "mlx5_bond_0,mlx5_bond_1,mlx5_bond_2,mlx5_bond_3"
- --cuda-graph-max-bs-decode
- "64"
- --max-running-requests
- "2048"
- --tp-size
- "16" # Size of Tensor Parallelism
- --dist-init-addr
- $(LWS_LEADER_ADDRESS):20102
- --nnodes
- $(LWS_GROUP_SIZE)
- --node-rank
- $(LWS_WORKER_INDEX)
- --trust-remote-code
- --ep-num-redundant-experts
- "32"
- --moe-dense-tp-size
- "1"
env:
- name: SGLANG_HACK_DEEPEP_NUM_SMS
value: "24"
- name: SGLANG_HACK_DEEPEP_NEW_MODE
value: "0"
- name: NVSHMEM_IB_TRAFFIC_CLASS
value: "16"
- name: NVSHMEM_IB_GID_INDEX
value: "3"
- name: NVSHMEM_ENABLE_NIC_PE_MAPPING
value: "1"
- name: NCCL_IB_QPS_PER_CONNECTION
value: "8"
- name: NCCL_IB_SPLIT_DATA_ON_QPS
value: "1"
- name: NCCL_NET_PLUGIN
value: "none"
- name: NCCL_IB_TC
value: "136"
- name: NCCL_MIN_NCHANNELS
value: "4"
- name: MC_TE_METRIC
value: "true"
- name: NCCL_IB_SL
value: "5"
- name: SGLANG_MOONCAKE_TRANS_THREAD
value: "16"
- name: SGLANG_ENABLE_JIT_DEEPGEMM
value: "1"
- name: NCCL_IB_HCA
value: ^=mlx5_0,mlx5_5,mlx5_6
- name: LWS_WORKER_INDEX
valueFrom:
fieldRef:
fieldPath: metadata.labels['leaderworkerset.sigs.k8s.io/worker-index']
image: lmsysorg/sglang:deepep
name: sglang-worker
ports:
- containerPort: 30001
resources:
limits:
nvidia.com/gpu: "8"
securityContext:
capabilities:
add:
- IPC_LOCK
privileged: true
volumeMounts:
- mountPath: /root/.cache
name: sgl-cache
- mountPath: /dev/shm
name: dshm
- mountPath: /work/models
name: model
- mountPath: /dev/infiniband
name: ib
- mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs
name: cf
dnsPolicy: ClusterFirstWithHostNet
hostIPC: true
hostNetwork: true
nodeSelector:
pd: "yes"
tolerations:
- key: pd
operator: Exists
- key: node-role
operator: Exists
volumes:
- hostPath:
path: /data1/sgl_cache1
type: DirectoryOrCreate
name: sgl-cache
- emptyDir:
medium: Memory
name: dshm
- hostPath:
path: /dev/infiniband
name: ib
- hostPath:
path: /data1/maas_hosted_models/models/DeepSeek-R1-0528/deepseek_r1_0528
name: model
- hostPath:
path: /data1/maas_hosted_models/models/fused_moe_triton/configs
name: cf
networkConfig:
subdomainPolicy: Shared
replicas: 1
rolloutStrategy:
rollingUpdateConfiguration:
maxSurge: 0
maxUnavailable: 1
type: RollingUpdate
startupPolicy: LeaderCreated
각각 별도로 실행해요:
kubectl apply -f p.yaml
kubectl apply -f d.yaml
이 시점에서 1P1D SGLang 엔진 부분의 배포가 완료됐어요.
사용자가 모델 API를 직접 체험할 수 있게 하려면 prefill과 decode 간 순차 호출을 처리할 로드 밸런서가 여전히 필요해요. 회사마다 LB 구현이 다르며, 커뮤니티가 곧 Rust로 작성된 새 LB 컴포넌트를 공식 릴리스할 거예요.
현재는 정적 K8S 서비스 + minilb 방식을 사용해 모델 API 호출을 구현해요.
Prefill과 Decode용 서비스 생성 (Creating Service for Prefill and Decode)
prefill k8s 서비스 생성
apiVersion: v1
kind: Service
metadata:
name: deepseekr10528-prefill-main
spec:
selector:
leaderworkerset.sigs.k8s.io/name: deepseekr10528-prefill-main
role: leader
ports:
- protocol: TCP
port: 30000
targetPort: 30000
kubectl apply -f p-svc.yaml 실행
decode k8s 서비스 생성
apiVersion: v1
kind: Service
metadata:
name: deepseekr10528-decode-main
spec:
selector:
leaderworkerset.sigs.k8s.io/name: deepseekr10528-decode-main
role: leader
ports:
- protocol: TCP
port: 30000
targetPort: 30000
kubectl apply -f d-svc.yaml 실행
minilb와 lb 서비스 배포
apiVersion: apps/v1
kind: Deployment
metadata:
name: deepseekr10528-lb-main
labels:
app: deepseekr10528-lb
spec:
replicas: 1
selector:
matchLabels:
app: deepseekr10528-lb
template:
metadata:
labels:
app: deepseekr10528-lb
spec:
nodeSelector:
pd: "yes"
tolerations:
- key: pd
operator: Exists
- key: node-role
operator: Exists
containers:
- name: sgl-minilb
image: lmsysorg/sglang:deepep
command:
- python
- -m
- sglang_router.launch_router
- --pd-disaggregation
- --prefill
- http://deepseekr10528-prefill-main:30000
- --decode
- http://deepseekr10528-decode-main:30000
- --host
- 0.0.0.0
- --port
- "8000"
ports:
- containerPort: 8000
***
apiVersion: v1
kind: Service
metadata:
name: deepseekr10528-lb-service
spec:
type: NodePort
selector:
app: deepseekr10528-lb
ports:
- protocol: TCP
port: 8000 # Service Port (In-Cluster)
targetPort: 8000 # Exposed Container
nodePort: 30800
kubectl apply -f lb.yaml 실행
모든 모델 배포 성공을 기다린 후 다음 출력을 얻을 수 있어요:
[root@ecs-001]# kubectl get po
deepseekr10528-decode-main-0 1/1 Running 0 74m
deepseekr10528-decode-main-0-1 1/1 Running 0 74m
deepseekr10528-lb-main-9c5dbfc57-6lcbd 1/1 Running 0 22m
deepseekr10528-prefill-main-0 1/1 Running 0 74m
deepseekr10528-prefill-main-0-1 1/1 Running 0 74m
이 시점에서 nodePort:30800을 선택해 접근해요:
curl -X POST "http://{nodePort}:30800/v1/chat/completions" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer ***" \
-d '{
"rid":"ccccdd",
"model": "r1",
"messages": [
{"role": "system", "content": "0: You are a helpful AI assistant"},
{"role": "user", "content": "你是谁？."}
],
"max_tokens":221
}'
FAQ
-
현재 배포 시작 파라미터는 모든 RDMA 시나리오와 완전히 호환되지 않을 수 있어요. 네트워크 환경에 따라 다른 RDMA NCCL 관련 환경 구성이 필요할 수 있어요.
-
EPLB의 일부 프리셋 최적화 구성은 여기서 사용되지 않아요. 필요에 따라 6017을 따라 조정할 수 있어요.