A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 23:45 UTC
[Take 2] Correctly kill vLLM processes after benchmarks (#21646)
Signed-off-by: Huy Do <huydhn@gmail.com>
Huy Do committed
Jul 26, 2025 at 06:06 UTC
e98def439cc89adba6d704083c5c78dd89b51b5f
3 files changed
+6
-1
.buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+2
-1
@@ -126,7 +126,8 @@ kill_gpu_processes() {
126
ps -aux
127
lsof -t -i:8000 | xargs -r kill -9
128
pgrep python3 | xargs -r kill -9
129
-
129
+ # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
130
+ pgrep VLLM | xargs -r kill -9
131
132
# wait until GPU memory usage smaller than 1GB
133
if command -v nvidia-smi; then
benchmarks/disagg_benchmarks/disagg_overhead_benchmark.sh
+2
@@ -12,6 +12,8 @@ kill_gpu_processes() {
12
# kill all processes on GPU.
13
pgrep pt_main_thread | xargs -r kill -9
14
pgrep python3 | xargs -r kill -9
15
+ # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
16
+ pgrep VLLM | xargs -r kill -9
17
sleep 10
18
19
# remove vllm config file
benchmarks/disagg_benchmarks/disagg_performance_benchmark.sh
+2
@@ -18,6 +18,8 @@ kill_gpu_processes() {
18
# kill all processes on GPU.
19
pgrep pt_main_thread | xargs -r kill -9
20
pgrep python3 | xargs -r kill -9
21
+ # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
22
+ pgrep VLLM | xargs -r kill -9
23
for port in 8000 8100 8200; do lsof -t -i:$port | xargs -r kill -9; done
24
sleep 1
25
}