Skip to content

[Fix] Failed to start vLLM service using multi-node launch scripts under CUDA data parallelism #670

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mag1c-h merged 1 commit into from
Jan 23, 2026
Merged

[Fix] Failed to start vLLM service using multi-node launch scripts under CUDA data parallelism #670

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions examples/deployments/scripts/vllm/common.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,13 +38,13 @@ ensure_ifconfig_installed() {

if command -v apt-get >/dev/null 2>&1; then
echo "Detected apt-get (Debian/Ubuntu). Installing net-tools..."
sudo apt-get update && sudo apt-get install -y net-tools
apt-get update && apt-get install -y net-tools
elif command -v yum >/dev/null 2>&1; then
echo "Detected yum (RHEL/CentOS). Installing net-tools..."
sudo yum install -y net-tools
yum install -y net-tools
elif command -v dnf >/dev/null 2>&1; then
echo "Detected dnf (Fedora). Installing net-tools..."
sudo dnf install -y net-tools
dnf install -y net-tools
else
echo "ERROR: No supported package manager (apt/yum/dnf) found."
echo "Please install 'net-tools' manually, 'ifconfig' is required to get network interface information."
Expand Down
6 changes: 5 additions & 1 deletion examples/deployments/scripts/vllm/run_vllm.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ start_server() {
echo "pp_size = $pp_size"
echo "enable_expert_parallel = $enable_expert_parallel"
echo "max_model_len = $max_model_len"
echo "max_num_batched_tokens = $max_num_batch_tokens"
echo "max_num_batched_tokens = $max_num_batched_tokens"
echo "max_num_seqs = $max_num_seqs"
echo "block_size = $block_size"
echo "gpu_memory_utilization = $gpu_memory_utilization"
Expand Down Expand Up @@ -69,6 +69,10 @@ start_server() {
fi

# --- Boolean flags ---
if [[ "$distributed_executor_backend" == "ray" ]] && [[ "$dp_size" -gt 1 ]]; then
CMD+=("--data-parallel-backend" "ray")
CMD+=("--data-parallel-size-local" "$((dp_size / node_num))");
fi
if [[ "$async_scheduling" == "true" ]]; then CMD+=("--async-scheduling"); fi
if [[ "$enable_expert_parallel" == "true" ]]; then CMD+=("--enable-expert-parallel"); fi
if [[ "$enable_prefix_caching" == "false" ]]; then CMD+=("--no-enable-prefix-caching"); fi
Expand Down
2 changes: 1 addition & 1 deletion examples/deployments/scripts/vllm/start_ray.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ set_node_env(){
export NCCL_SOCKET_IFNAME="$IFACE"
export GLOO_SOCKET_IFNAME="$IFACE"
export TP_SOCKET_IFNAME="$IFACE"
export NUM_GPUS=$((tp_size / node_num))
export NUM_GPUS=$((tp_size * dp_size * pp_size / node_num))

echo ""
echo "===== ray startup configuration ======"
Expand Down
Loading