.env configuration support for launch-cluster.sh

This commit is contained in:
Eugene Rakhmatulin
2026-03-25 14:18:00 -07:00
parent c4b078b868
commit ad2cd3373f
3 changed files with 548 additions and 231 deletions

29
.env.example Normal file
View File

@@ -0,0 +1,29 @@
# Example .env configuration file for spark-vllm-docker
# Copy this file to .env and customize for your environment
# Cluster configuration
# CLUSTER_NODES: Comma-separated list of node IPs (first node is the head node)
CLUSTER_NODES="192.168.1.1,192.168.1.2,192.168.1.3"
# ETH_IF: Ethernet interface name (optional, auto-detected if not specified)
ETH_IF="eth0"
# IB_IF: InfiniBand interface name (optional, auto-detected if not specified)
IB_IF="ib0"
# MASTER_PORT: Port for cluster coordination (default: 29501)
MASTER_PORT="29501"
# CONTAINER_NAME: Container name (default: vllm_node)
CONTAINER_NAME="vllm_node"
# Container environment variables
# Any variable starting with CONTAINER_ will be converted to -e flags
# Example: CONTAINER_NCCL_DEBUG=INFO becomes -e NCCL_DEBUG=INFO
CONTAINER_NCCL_DEBUG="INFO"
CONTAINER_HF_TOKEN="your_huggingface_token_here"
CONTAINER_NCCL_IGNORE_CPU_AFFINITY="1"
# Additional container environment variables
# CONTAINER_MAX_JOBS="16"
# CONTAINER_CUDA_VISIBLE_DEVICES="0,1"