diff --git a/.github/workflows/buildkit-dockerfile-test.yaml b/.github/workflows/buildkit-dockerfile-test.yaml
new file mode 100644
index 000000000..973c38730
--- /dev/null
+++ b/.github/workflows/buildkit-dockerfile-test.yaml
@@ -0,0 +1,86 @@
+name: buildkit-dockerfile-test
+
+on:
+  pull_request:
+    paths:
+      - '.github/workflows/buildkit-dockerfile-test.yaml'
+      - 'Dockerfile*'
+  workflow_dispatch:
+
+jobs:
+  build-test:
+    name: "${{ matrix.dockerfile }} (${{ matrix.runner_info.arch }})"
+    runs-on: ${{ matrix.runner_info.runner }}
+    container:
+      image: swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/cann:9.0.0-a3-ubuntu22.04-py3.12
+
+    strategy:
+      fail-fast: false
+      matrix:
+        dockerfile:
+          - Dockerfile
+          - Dockerfile.310p
+          - Dockerfile.310p.openEuler
+          - Dockerfile.a3
+          - Dockerfile.a3.openEuler
+          - Dockerfile.test
+        runner_info:
+          - {runner: linux-aarch64-cpu-4-buildkit-gy006, arch: arm64}
+          - {runner: linux-amd64-cpu-4-buildkit-gy006, arch: amd64}
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+          persist-credentials: false
+
+      - name: Determine SOC_VERSION
+        id: vars
+        run: |
+          case "${{ matrix.dockerfile }}" in
+            Dockerfile|Dockerfile.openEuler)
+              echo "soc_version=ascend910b1" >> $GITHUB_OUTPUT
+              ;;
+            Dockerfile.a3|Dockerfile.a3.openEuler)
+              echo "soc_version=ascend910_9391" >> $GITHUB_OUTPUT
+              ;;
+            Dockerfile.a5|Dockerfile.a5.openEuler)
+              echo "soc_version=ascend950dt_9582" >> $GITHUB_OUTPUT
+              ;;
+            Dockerfile.310p|Dockerfile.310p.openEuler)
+              echo "soc_version=ascend310p1" >> $GITHUB_OUTPUT
+              ;;
+          esac
+
+      - name: Verify buildctl environment
+        run: |
+          echo "BUILDKITD_ADDR=$BUILDKITD_ADDR"
+          test -n "$BUILDKITD_ADDR" || (echo "ERROR: BUILDKITD_ADDR not set" && exit 1)
+          test -f "${DOCKER_CONFIG}/ca.pem" || (echo "ERROR: ca.pem not found" && exit 1)
+          test -f "${DOCKER_CONFIG}/cert.pem" || (echo "ERROR: cert.pem not found" && exit 1)
+          test -f "${DOCKER_CONFIG}/key.pem" || (echo "ERROR: key.pem not found" && exit 1)
+          buildctl --version
+
+      - name: Build image with buildctl
+        run: |
+          buildctl \
+            --addr="${BUILDKITD_ADDR}" \
+            --tlscacert="${DOCKER_CONFIG}/ca.pem" \
+            --tlscert="${DOCKER_CONFIG}/cert.pem" \
+            --tlskey="${DOCKER_CONFIG}/key.pem" \
+            build \
+            --progress=plain \
+            --frontend dockerfile.v0 \
+            --local context=. \
+            --local dockerfile=. \
+            --opt filename=${{ matrix.dockerfile }} \
+            --opt build-arg:APTMIRROR=http://cache-service.nginx-pypi-cache.svc.cluster.local:8081 \
+            --opt build-arg:PIP_INDEX_URL=http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple \
+            --opt build-arg:VLLM_ASCEND_IMAGE=swr.cn-southwest-2.myhuaweicloud.com/base_image/ascend-ci/vllm-ascend/vllm-ascend-ci \
+            --opt build-arg:SOC_VERSION=${{ steps.vars.outputs.soc_version }} \
+            --opt build-arg:COMPILE_CUSTOM_KERNELS=0 \
+            --secret id=dockerconfig,src=/home/user/.docker/config.json \
+            --import-cache type=registry,ref=swr.cn-southwest-2.myhuaweicloud.com/modelfoundry/buildkit-cache:${{ matrix.dockerfile }}-${{ matrix.runner_info.arch }} \
+            --export-cache type=registry,ref=swr.cn-southwest-2.myhuaweicloud.com/modelfoundry/buildkit-cache:${{ matrix.dockerfile }}-${{ matrix.runner_info.arch }},mode=max \
+            --output type=image,name=swr.cn-southwest-2.myhuaweicloud.com/modelfoundry/test-buildkit:${{ matrix.dockerfile }}-${{ matrix.runner_info.arch }}-${{ github.sha }},push=true
diff --git a/Dockerfile b/Dockerfile
index f923a7a5f..edb918d21 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -17,23 +17,31 @@
 
 FROM quay.io/ascend/cann:9.0.1-910b-ubuntu22.04-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG MOONCAKE_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
+ARG APTMIRROR
 
 WORKDIR /workspace
 
 # Install clang-15 (for triton-ascend) and Mooncake
 ARG MOONCAKE_TAG=0.3.11.post1
-RUN apt-get update -y && \
+RUN if [ -n "$APTMIRROR" ]; then \
+        sed -Ei "s@(ports|archive).ubuntu.com@${APTMIRROR#http://}@g" /etc/apt/sources.list; \
+    fi && \
+    apt-get update -y && \
     apt-get install -y git vim wget net-tools gcc g++ cmake numactl libnuma-dev libibverbs-dev libjemalloc2 libhiredis-dev clang-15 && \
     update-alternatives --install /usr/bin/clang clang /usr/bin/clang-15 20 && \
     update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-15 20 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url https://mirrors.aliyun.com/pypi/web/simple && \
+    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url ${MOONCAKE_INDEX_URL} && \
     rm -rf /var/cache/apt/* && \
     rm -rf /var/lib/apt/lists/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -49,7 +57,7 @@ RUN if [ -n "$VLLM_COMMIT" ]; then \
       git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
     fi
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -62,13 +70,13 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     export VLLM_BATCH_INVARIANT=1 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton triton-ascend && \
-    python3 -m pip install triton-ascend==3.2.1 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi && \
+    python3 -m pip install triton-ascend==3.2.1 --extra-index-url ${ASCEND_INDEX_URL} && \
     python3 -m pip cache purge
 
 # Append `libascend_hal.so` path (devlib) to LD_LIBRARY_PATH
diff --git a/Dockerfile.310p b/Dockerfile.310p
index 2f97a47dd..5924de6e5 100644
--- a/Dockerfile.310p
+++ b/Dockerfile.310p
@@ -17,17 +17,24 @@
 
 FROM quay.io/ascend/cann:9.1.0-beta.1-310p-ubuntu22.04-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
+ARG APTMIRROR
 
 WORKDIR /workspace
 
-RUN apt-get update -y && \
+RUN if [ -n "$APTMIRROR" ]; then \
+        sed -Ei "s@(ports|archive).ubuntu.com@${APTMIRROR#http://}@g" /etc/apt/sources.list; \
+    fi && \
+    apt-get update -y && \
     apt-get install -y python3-pip git vim wget net-tools gcc g++ cmake numactl libnuma-dev libjemalloc2 pciutils && \
     rm -rf /var/cache/apt/* && \
     rm -rf /var/lib/apt/lists/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -43,7 +50,7 @@ RUN if [ -n "$VLLM_COMMIT" ]; then \
       git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
     fi
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -56,10 +63,10 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton-ascend triton && \
     python3 -m pip cache purge
 
diff --git a/Dockerfile.310p.openEuler b/Dockerfile.310p.openEuler
index a7ba0336a..4cbcfb3bd 100644
--- a/Dockerfile.310p.openEuler
+++ b/Dockerfile.310p.openEuler
@@ -17,7 +17,9 @@
 
 FROM quay.io/ascend/cann:9.1.0-beta.1-310p-openeuler24.03-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
 
 WORKDIR /workspace
 
@@ -27,6 +29,7 @@ RUN yum update -y && \
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -42,7 +45,7 @@ RUN if [ -n "$VLLM_COMMIT" ]; then \
       git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
     fi
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -54,10 +57,10 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton-ascend triton && \
     python3 -m pip cache purge
 
diff --git a/Dockerfile.a3 b/Dockerfile.a3
index 307fbac6e..244852dcf 100644
--- a/Dockerfile.a3
+++ b/Dockerfile.a3
@@ -17,7 +17,11 @@
 
 FROM quay.io/ascend/cann:9.0.1-a3-ubuntu22.04-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG MOONCAKE_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
+ARG APTMIRROR
 
 ENV DEBIAN_FRONTEND=noninteractive
 
@@ -25,17 +29,21 @@ WORKDIR /workspace
 
 # Install clang-15 (for triton-ascend) and Mooncake
 ARG MOONCAKE_TAG=0.3.11.post1
-RUN apt-get update -y && \
+RUN if [ -n "$APTMIRROR" ]; then \
+        sed -Ei "s@(ports|archive).ubuntu.com@${APTMIRROR#http://}@g" /etc/apt/sources.list; \
+    fi && \
+    apt-get update -y && \
     apt-get install -y git vim wget net-tools gcc g++ cmake numactl libnuma-dev libibverbs-dev libjemalloc2 libhiredis-dev clang-15 && \
     update-alternatives --install /usr/bin/clang clang /usr/bin/clang-15 20 && \
     update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-15 20 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url https://mirrors.aliyun.com/pypi/web/simple && \
+    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url ${MOONCAKE_INDEX_URL} && \
     rm -rf /var/cache/apt/* && \
     rm -rf /var/lib/apt/lists/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -51,7 +59,7 @@ RUN if [ -n "$VLLM_COMMIT" ]; then \
       git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
     fi
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -64,13 +72,13 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     export VLLM_BATCH_INVARIANT=1 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton triton-ascend && \
-    python3 -m pip install triton-ascend==3.2.1 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi && \
+    python3 -m pip install triton-ascend==3.2.1 --extra-index-url ${ASCEND_INDEX_URL} && \
     python3 -m pip cache purge
 
 # Append `libascend_hal.so` path (devlib) to LD_LIBRARY_PATH
diff --git a/Dockerfile.a3.openEuler b/Dockerfile.a3.openEuler
index d8b2a7da3..dd5836759 100644
--- a/Dockerfile.a3.openEuler
+++ b/Dockerfile.a3.openEuler
@@ -17,7 +17,10 @@
 
 FROM quay.io/ascend/cann:9.0.1-a3-openeuler24.03-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG MOONCAKE_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
 
 WORKDIR /workspace
 
@@ -28,11 +31,12 @@ ARG MOONCAKE_TAG=0.3.11.post1
 RUN yum update -y && \
     yum install -y git vim wget net-tools gcc gcc-c++ make cmake numactl numactl-devel libibverbs-devel jemalloc hiredis-devel clang patch && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url https://mirrors.aliyun.com/pypi/web/simple && \
+    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url ${MOONCAKE_INDEX_URL} && \
     rm -rf /var/cache/yum/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -48,7 +52,7 @@ RUN if [ -n "$VLLM_COMMIT" ]; then \
       git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
     fi
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -60,13 +64,13 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     export VLLM_BATCH_INVARIANT=1 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton triton-ascend && \
-    python3 -m pip install triton-ascend==3.2.1 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi && \
+    python3 -m pip install triton-ascend==3.2.1 --extra-index-url ${ASCEND_INDEX_URL} && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc
diff --git a/Dockerfile.a5 b/Dockerfile.a5
index 3b1bcb9fa..a736556ec 100644
--- a/Dockerfile.a5
+++ b/Dockerfile.a5
@@ -17,7 +17,11 @@
 
 FROM quay.io/ascend/cann:9.0.1-950-ubuntu22.04-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG MOONCAKE_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
+ARG APTMIRROR
 
 ENV DEBIAN_FRONTEND=noninteractive
 
@@ -25,17 +29,21 @@ WORKDIR /workspace
 
 # Install clang-15 (for triton-ascend) and Mooncake
 ARG MOONCAKE_TAG=0.3.11.post1
-RUN apt-get update -y && \
+RUN if [ -n "$APTMIRROR" ]; then \
+        sed -Ei "s@(ports|archive).ubuntu.com@${APTMIRROR#http://}@g" /etc/apt/sources.list; \
+    fi && \
+    apt-get update -y && \
     apt-get install -y git vim wget net-tools gcc g++ cmake numactl libnuma-dev libibverbs-dev libjemalloc2 libhiredis-dev clang-15 && \
     update-alternatives --install /usr/bin/clang clang /usr/bin/clang-15 20 && \
     update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-15 20 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url https://mirrors.aliyun.com/pypi/web/simple && \
+    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url ${MOONCAKE_INDEX_URL} && \
     rm -rf /var/cache/apt/* && \
     rm -rf /var/lib/apt/lists/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -44,7 +52,7 @@ ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
 ARG VLLM_TAG=v0.25.1
 RUN git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -57,12 +65,12 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton triton-ascend && \
-    python3 -m pip install triton-ascend==3.2.1 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi && \
+    python3 -m pip install triton-ascend==3.2.1 --extra-index-url ${ASCEND_INDEX_URL} && \
     python3 -m pip cache purge
 
 # Append `libascend_hal.so` path (devlib) to LD_LIBRARY_PATH
diff --git a/Dockerfile.a5.openEuler b/Dockerfile.a5.openEuler
index 73cd94214..246b373c7 100644
--- a/Dockerfile.a5.openEuler
+++ b/Dockerfile.a5.openEuler
@@ -17,7 +17,10 @@
 
 FROM quay.io/ascend/cann:9.0.1-950-openeuler24.03-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG MOONCAKE_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
 
 WORKDIR /workspace
 
@@ -28,11 +31,12 @@ ARG MOONCAKE_TAG=0.3.11.post1
 RUN yum update -y && \
     yum install -y git vim wget net-tools gcc gcc-c++ make cmake numactl numactl-devel libibverbs-devel jemalloc hiredis-devel clang patch && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url https://mirrors.aliyun.com/pypi/web/simple && \
+    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url ${MOONCAKE_INDEX_URL} && \
     rm -rf /var/cache/yum/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -41,7 +45,7 @@ ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
 ARG VLLM_TAG=v0.25.1
 RUN git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -53,12 +57,12 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton triton-ascend && \
-    python3 -m pip install triton-ascend==3.2.1 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi && \
+    python3 -m pip install triton-ascend==3.2.1 --extra-index-url ${ASCEND_INDEX_URL} && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc
diff --git a/Dockerfile.openEuler b/Dockerfile.openEuler
index 25c1f1fe7..999be8af3 100644
--- a/Dockerfile.openEuler
+++ b/Dockerfile.openEuler
@@ -17,7 +17,10 @@
 
 FROM quay.io/ascend/cann:9.0.1-910b-openeuler24.03-py3.12
 
-ARG PIP_INDEX_URL="https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple"
+ARG PIP_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG MOONCAKE_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/pypi/simple"
+ARG ASCEND_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi"
+ARG PYTORCH_INDEX_URL="http://cache-service.nginx-pypi-cache.svc.cluster.local/whl/cpu"
 
 WORKDIR /workspace
 
@@ -28,11 +31,12 @@ ARG MOONCAKE_TAG=0.3.11.post1
 RUN yum update -y && \
     yum install -y git vim wget net-tools gcc gcc-c++ make cmake numactl numactl-devel libibverbs-devel jemalloc hiredis-devel clang patch && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url https://mirrors.aliyun.com/pypi/web/simple && \
+    python3 -m pip install mooncake-transfer-engine-npu==${MOONCAKE_TAG} --extra-index-url ${MOONCAKE_INDEX_URL} && \
     rm -rf /var/cache/yum/*
 
 # Install modelscope (for fast download) and ray (for multinode)
 RUN pip config set global.index-url ${PIP_INDEX_URL} && \
+    pip config set global.trusted-host cache-service.nginx-pypi-cache.svc.cluster.local && \
     python3 -m pip install modelscope 'ray>=2.47.1,<=2.48.0' 'protobuf>3.20.0' && \
     python3 -m pip cache purge
 
@@ -48,7 +52,7 @@ RUN if [ -n "$VLLM_COMMIT" ]; then \
       git clone --depth 1 -b $VLLM_TAG $VLLM_REPO /vllm-workspace/vllm; \
     fi
 # In x86, triton will be installed by vllm. But in Ascend, triton doesn't work correctly. we need to uninstall it.
-RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index https://download.pytorch.org/whl/cpu/ && \
+RUN VLLM_TARGET_DEVICE="empty" python3 -m pip install -e /vllm-workspace/vllm/[audio] --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton && \
     python3 -m pip cache purge
 
@@ -60,13 +64,13 @@ ENV SOC_VERSION=$SOC_VERSION \
     OMP_NUM_THREADS=1
 COPY . /vllm-workspace/vllm-ascend/
 
-RUN export PIP_EXTRA_INDEX_URL="https://mirrors.huaweicloud.com/ascend/repos/pypi" && \
+RUN export PIP_EXTRA_INDEX_URL="${ASCEND_INDEX_URL}" && \
     export VLLM_BATCH_INVARIANT=1 && \
     source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
     source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/ && \
+    python3 -m pip install -e /vllm-workspace/vllm-ascend/ --extra-index ${PYTORCH_INDEX_URL} && \
     python3 -m pip uninstall -y triton triton-ascend && \
-    python3 -m pip install triton-ascend==3.2.1 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi && \
+    python3 -m pip install triton-ascend==3.2.1 --extra-index-url ${ASCEND_INDEX_URL} && \
     python3 -m pip cache purge
 
 RUN echo "export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD" >> ~/.bashrc