Avoid stale exact-trace HTTP connections
This commit is contained in:
@@ -171,7 +171,7 @@ async def replay(args: argparse.Namespace, rows: list[dict[str, Any]]) -> list[d
|
|||||||
import aiohttp
|
import aiohttp
|
||||||
|
|
||||||
timeout = aiohttp.ClientTimeout(total=args.timeout_seconds)
|
timeout = aiohttp.ClientTimeout(total=args.timeout_seconds)
|
||||||
connector = aiohttp.TCPConnector(limit=0, ttl_dns_cache=300)
|
connector = aiohttp.TCPConnector(limit=0, ttl_dns_cache=300, force_close=True)
|
||||||
benchmark_start = asyncio.get_running_loop().time() + 2.0
|
benchmark_start = asyncio.get_running_loop().time() + 2.0
|
||||||
async with aiohttp.ClientSession(
|
async with aiohttp.ClientSession(
|
||||||
base_url=f"http://{args.host}:{args.port}",
|
base_url=f"http://{args.host}:{args.port}",
|
||||||
@@ -229,6 +229,7 @@ def main() -> None:
|
|||||||
"arrival": "original_trace_timestamp_and_order",
|
"arrival": "original_trace_timestamp_and_order",
|
||||||
"input_output_prompt": "exact_source_values",
|
"input_output_prompt": "exact_source_values",
|
||||||
"served_model_alias": args.served_model,
|
"served_model_alias": args.served_model,
|
||||||
|
"http_connection_reuse": False,
|
||||||
"ttft_slo": "1000ms + 1000ms * input_tokens / 8000",
|
"ttft_slo": "1000ms + 1000ms * input_tokens / 8000",
|
||||||
"tpot_slo_ms": args.tpot_slo_ms,
|
"tpot_slo_ms": args.tpot_slo_ms,
|
||||||
"target_pass_rate": TARGET_PASS_RATE,
|
"target_pass_rate": TARGET_PASS_RATE,
|
||||||
|
|||||||
Reference in New Issue
Block a user