Remove redundant --http-port flag from frontend launch commands

keivenchang · keivenchang · commit b0a5e587d8c3 · 2025-11-05T09:12:20.000-08:00
dynamo.frontend already reads DYN_HTTP_PORT env var and defaults to 8000,
so explicit --http-port=${DYN_HTTP_PORT:-8000} flag is redundant. Add
comment to clarify this behavior for maintainability.

Signed-off-by: Keiven Chang &lt;keivenchang@users.noreply.github.com&gt;
diff --git a/examples/backends/sglang/launch/agg.sh b/examples/backends/sglang/launch/agg.sh
@@ -13,7 +13,8 @@ trap cleanup EXIT INT TERM
 
 
 # run ingress
-python3 -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run worker with metrics enabled
diff --git a/examples/backends/sglang/launch/agg_embed.sh b/examples/backends/sglang/launch/agg_embed.sh
@@ -13,7 +13,8 @@ trap cleanup EXIT INT TERM
 
 
 # run ingress
-python3 -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run worker
diff --git a/examples/backends/sglang/launch/agg_router.sh b/examples/backends/sglang/launch/agg_router.sh
@@ -13,7 +13,8 @@ trap cleanup EXIT INT TERM
 
 
 # run ingress
-python -m dynamo.frontend --router-mode kv --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend --router-mode kv &
 DYNAMO_PID=$!
 
 # run worker
diff --git a/examples/backends/sglang/launch/disagg.sh b/examples/backends/sglang/launch/disagg.sh
@@ -13,7 +13,8 @@ trap cleanup EXIT INT TERM
 
 
 # run ingress
-python3 -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run prefill worker
diff --git a/examples/backends/sglang/launch/disagg_dp_attn.sh b/examples/backends/sglang/launch/disagg_dp_attn.sh
@@ -13,7 +13,8 @@ trap cleanup EXIT INT TERM
 
 
 # run ingress
-python3 -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run prefill worker
diff --git a/examples/backends/sglang/launch/disagg_same_gpu.sh b/examples/backends/sglang/launch/disagg_same_gpu.sh
@@ -37,7 +37,8 @@ trap cleanup EXIT INT TERM
 
 
 # run ingress with KV router mode for disaggregated setup
-python3 -m dynamo.frontend --router-mode kv --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend --router-mode kv &
 DYNAMO_PID=$!
 
 # run prefill worker with metrics on port 8081
diff --git a/examples/backends/sglang/launch/multimodal_agg.sh b/examples/backends/sglang/launch/multimodal_agg.sh
@@ -60,7 +60,8 @@ if [[ -n "$SERVED_MODEL_NAME" ]]; then
 fi
 
 # run ingress
-python3 -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run SGLang multimodal processor
diff --git a/examples/backends/sglang/launch/multimodal_disagg.sh b/examples/backends/sglang/launch/multimodal_disagg.sh
@@ -60,7 +60,8 @@ if [[ -n "$SERVED_MODEL_NAME" ]]; then
 fi
 
 # run ingress
-python3 -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run SGLang multimodal processor
diff --git a/examples/backends/trtllm/launch/agg.sh b/examples/backends/trtllm/launch/agg.sh
@@ -22,7 +22,8 @@ trap cleanup EXIT INT TERM
 
 
 # run frontend
-python3 -m dynamo.frontend --http-port ${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run worker
diff --git a/examples/backends/trtllm/launch/agg_metrics.sh b/examples/backends/trtllm/launch/agg_metrics.sh
@@ -19,7 +19,8 @@ cleanup() {
 trap cleanup EXIT INT TERM
 
 # Run frontend
-python3 -m dynamo.frontend --http-port ${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # Run worker
diff --git a/examples/backends/trtllm/launch/disagg.sh b/examples/backends/trtllm/launch/disagg.sh
@@ -26,7 +26,8 @@ trap cleanup EXIT INT TERM
 
 
 # run frontend
-python3 -m dynamo.frontend --http-port ${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run prefill worker
diff --git a/examples/backends/trtllm/launch/disagg_same_gpu.sh b/examples/backends/trtllm/launch/disagg_same_gpu.sh
@@ -49,7 +49,8 @@ trap cleanup EXIT INT TERM
 
 
 # run frontend
-python3 -m dynamo.frontend --http-port ${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run prefill worker (shares GPU with decode)
diff --git a/examples/backends/trtllm/launch/epd_disagg.sh b/examples/backends/trtllm/launch/epd_disagg.sh
@@ -29,7 +29,8 @@ trap cleanup EXIT INT TERM
 
 
 # run frontend
-python3 -m dynamo.frontend --http-port ${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python3 -m dynamo.frontend &
 DYNAMO_PID=$!
 
 # run encode worker
diff --git a/examples/backends/trtllm/performance_sweeps/scripts/start_frontend.sh b/examples/backends/trtllm/performance_sweeps/scripts/start_frontend.sh
@@ -20,6 +20,7 @@ etcd --listen-client-urls http://0.0.0.0:2379 --advertise-client-urls http://0.0
 sleep 2
 
 # Start OpenAI Frontend which will dynamically discover workers when they startup
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
 # NOTE: This is a blocking call.
-python3 -m dynamo.frontend --http-port ${DYN_HTTP_PORT:-8000}
+python3 -m dynamo.frontend
 
diff --git a/examples/backends/vllm/launch/agg.sh b/examples/backends/vllm/launch/agg.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # run worker
 # --enforce-eager is added for quick deployment. for production use, need to remove this flag
diff --git a/examples/backends/vllm/launch/agg_kvbm.sh b/examples/backends/vllm/launch/agg_kvbm.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # run worker with KVBM enabled
 # NOTE: remove --enforce-eager for production use
diff --git a/examples/backends/vllm/launch/agg_lmcache.sh b/examples/backends/vllm/launch/agg_lmcache.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # run worker with LMCache enabled
 ENABLE_LMCACHE=1 \
diff --git a/examples/backends/vllm/launch/agg_multimodal.sh b/examples/backends/vllm/launch/agg_multimodal.sh
@@ -53,7 +53,8 @@ else
 fi
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # To make Qwen2.5-VL fit in A100 40GB, set the following extra arguments
 EXTRA_ARGS=""
diff --git a/examples/backends/vllm/launch/agg_multimodal_llama.sh b/examples/backends/vllm/launch/agg_multimodal_llama.sh
@@ -8,7 +8,8 @@ trap 'echo Cleaning up...; kill 0' EXIT
 MODEL_NAME="meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # run processor
 python -m dynamo.vllm --multimodal-processor --model $MODEL_NAME --mm-prompt-template "<|image|>\n<prompt>" &
diff --git a/examples/backends/vllm/launch/dep.sh b/examples/backends/vllm/launch/dep.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress
-python -m dynamo.frontend --router-mode kv --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend --router-mode kv &
 
 # Data Parallel Attention / Expert Parallelism
 # Routing to DP workers managed by Dynamo
diff --git a/examples/backends/vllm/launch/disagg.sh b/examples/backends/vllm/launch/disagg.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # --enforce-eager is added for quick deployment. for production use, need to remove this flag
 CUDA_VISIBLE_DEVICES=0 python3 -m dynamo.vllm --model Qwen/Qwen3-0.6B --enforce-eager &
diff --git a/examples/backends/vllm/launch/disagg_kvbm.sh b/examples/backends/vllm/launch/disagg_kvbm.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress
-python -m dynamo.frontend --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend &
 
 # run decode worker on GPU 0, without enabling KVBM
 # NOTE: remove --enforce-eager for production use
diff --git a/examples/backends/vllm/launch/disagg_kvbm_2p2d.sh b/examples/backends/vllm/launch/disagg_kvbm_2p2d.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress with KV router
-python -m dynamo.frontend --router-mode kv --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend --router-mode kv &
 
 # run decode workers on GPU 0 and 1, without enabling KVBM
 # NOTE: remove --enforce-eager for production use
diff --git a/examples/backends/vllm/launch/disagg_lmcache.sh b/examples/backends/vllm/launch/disagg_lmcache.sh
@@ -5,7 +5,8 @@ set -e
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress with KV router
-python -m dynamo.frontend --router-mode kv --http-port=${DYN_HTTP_PORT:-8000} &
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
+python -m dynamo.frontend --router-mode kv &
 
 # run decode worker on GPU 0, without enabling LMCache
 CUDA_VISIBLE_DEVICES=0 python3 -m dynamo.vllm --model Qwen/Qwen3-0.6B &
diff --git a/examples/backends/vllm/launch/dsr1_dep.sh b/examples/backends/vllm/launch/dsr1_dep.sh
@@ -82,8 +82,9 @@ echo "  Model name: $MODEL"
 trap 'echo Cleaning up...; kill 0' EXIT
 
 # run ingress if it's node 0
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
 if [ $NODE_RANK -eq 0 ]; then
-    DYN_LOG=debug python -m dynamo.frontend --router-mode kv --http-port=${DYN_HTTP_PORT:-8000} 2>&1 | tee $LOG_DIR/dsr1_dep_ingress.log &
+    DYN_LOG=debug python -m dynamo.frontend --router-mode kv 2>&1 | tee $LOG_DIR/dsr1_dep_ingress.log &
 fi
 
 mkdir -p $LOG_DIR
diff --git a/examples/basics/multinode/trtllm/start_frontend_services.sh b/examples/basics/multinode/trtllm/start_frontend_services.sh
@@ -12,5 +12,6 @@ etcd --listen-client-urls http://0.0.0.0:2379 --advertise-client-urls http://0.0
 sleep 3
 
 # Start OpenAI Frontend which will dynamically discover workers when they startup
+# DYN_HTTP_PORT env var is read by dynamo.frontend (defaults to 8000 if not set)
 # NOTE: This is a blocking call.
-python3 -m dynamo.frontend --http-port 8000
+python3 -m dynamo.frontend
