EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[bugfix] Fix 2 minor bugs in the hicache storage layer (#8404)

Browse Source
This commit is contained in:
yi wang
2025-07-31 13:47:14 +08:00
committed by GitHub
parent 43118f5f2a
commit 5963e50503
2 changed files with 6 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
examples/monitoring/grafana/dashboards/json/sglang-dashboard.json
View File

@@ -147,7 +147,7 @@
"hide": false, "hide": false,
"includeNullMetadata": true, "includeNullMetadata": true,
"instant": false, "instant": false,
"legendFormat": "P95", "legendFormat": "P50",
"range": true, "range": true,
"refId": "C", "refId": "C",
"useBackend": false "useBackend": false
@@ -164,7 +164,7 @@
"hide": false, "hide": false,
"includeNullMetadata": true, "includeNullMetadata": true,
"instant": false, "instant": false,
"legendFormat": "P50", "legendFormat": "Avg",
"range": true, "range": true,
"refId": "D", "refId": "D",
"useBackend": false "useBackend": false
@@ -393,7 +393,7 @@
"hide": false, "hide": false,
"includeNullMetadata": true, "includeNullMetadata": true,
"instant": false, "instant": false,
"legendFormat": "P95", "legendFormat": "P50",
"range": true, "range": true,
"refId": "C", "refId": "C",
"useBackend": false "useBackend": false
@@ -405,12 +405,12 @@
}, },
"disableTextWrap": false, "disableTextWrap": false,
"editorMode": "code", "editorMode": "code",
"expr": "avg(rate(sglang:time_to_first_token_seconds_bucket[$__rate_interval]) / rate(sglang:time_to_first_token_seconds_bucket[$__rate_interval]))\r\n", "expr": "avg(rate(sglang:time_to_first_token_seconds_sum[$__rate_interval]) / rate(sglang:time_to_first_token_seconds_count[$__rate_interval]))\r\n",
"fullMetaSearch": false, "fullMetaSearch": false,
"hide": false, "hide": false,
"includeNullMetadata": true, "includeNullMetadata": true,
"instant": false, "instant": false,
"legendFormat": "P50", "legendFormat": "Avg",
"range": true, "range": true,
"refId": "D", "refId": "D",
"useBackend": false "useBackend": false

1
python/sglang/srt/mem_cache/hiradix_cache.py
View File

@@ -113,6 +113,7 @@ class HiRadixCache(RadixCache):
) )
if host_indices is not None: if host_indices is not None:
node.host_value = host_indices node.host_value = host_indices
assert len(node.host_value) > 0
self.ongoing_write_through[node.id] = node self.ongoing_write_through[node.id] = node
if not write_back: if not write_back:
# no need to lock nodes if write back # no need to lock nodes if write back