Fix PD disaggregation warmup: set request_name and improve error logging (#25542)

This commit is contained in:
Lianmin Zheng
2026-05-18 09:49:56 -07:00
committed by GitHub
parent 3e2a109636
commit f21fe6ad4d
+4 -3
View File
@@ -1965,6 +1965,7 @@ def _execute_server_warmup(server_args: ServerArgs):
else: else:
logger.info(f"Start of pd disaggregation warmup ...") logger.info(f"Start of pd disaggregation warmup ...")
request_name = "/generate"
json_data = { json_data = {
"sampling_params": { "sampling_params": {
"temperature": 0.0, "temperature": 0.0,
@@ -1997,9 +1998,9 @@ def _execute_server_warmup(server_args: ServerArgs):
_global_state.tokenizer_manager.server_status = ServerStatus.Up _global_state.tokenizer_manager.server_status = ServerStatus.Up
else: else:
logger.info( logger.info(
"Prefill disaggregation mode warm Up Failed, status code: {}".format( "Disaggregation warmup failed (mode=%s), status code: %s",
res.status_code server_args.disaggregation_mode,
) res.status_code,
) )
_global_state.tokenizer_manager.server_status = ServerStatus.UnHealthy _global_state.tokenizer_manager.server_status = ServerStatus.UnHealthy