minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Fix corrupted JSONL metrics file due to concurrent writes (#19011)

Browse Source
This commit is contained in:
Talor Abramovich
2026-02-22 20:33:27 +08:00
committed by GitHub
parent 1f7a813051
commit c6a99e43b9
1 changed files with 12 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/request_metrics_exporter.py
+12 -10
View File
@@ -87,6 +87,7 @@ class FileRequestMetricsExporter(RequestMetricsExporter):
# File handler state management # File handler state management
self._current_file_handler = None self._current_file_handler = None
self._current_file_lock = asyncio.Lock()
self._current_hour_suffix = None self._current_hour_suffix = None
def _ensure_file_handler(self, hour_suffix: str): def _ensure_file_handler(self, hour_suffix: str):
@@ -135,20 +136,21 @@ class FileRequestMetricsExporter(RequestMetricsExporter):
current_time = datetime.now() current_time = datetime.now()
hour_suffix = current_time.strftime("%Y%m%d_%H") hour_suffix = current_time.strftime("%Y%m%d_%H")
# Ensure correct file handler is open for current hour async with self._current_file_lock:
self._ensure_file_handler(hour_suffix) # Ensure correct file handler is open for current hour
self._ensure_file_handler(hour_suffix)
if self._current_file_handler is None: if self._current_file_handler is None:
return return
metrics_data = self._format_output_data(obj, out_dict) metrics_data = self._format_output_data(obj, out_dict)
def write_file(): def write_file():
json.dump(metrics_data, self._current_file_handler) json.dump(metrics_data, self._current_file_handler)
self._current_file_handler.write("\n") self._current_file_handler.write("\n")
self._current_file_handler.flush() self._current_file_handler.flush()
await asyncio.to_thread(write_file) await asyncio.to_thread(write_file)
except Exception as e: except Exception as e:
logger.exception(f"Failed to write perf metrics to file: {e}") logger.exception(f"Failed to write perf metrics to file: {e}")