datacoord remote_chunk_manager
walk through object failed
#41022
-
Occasionally, we see a warning log like the one below. Is it okay to ignore it? Could this warning indicate a potential issue with Milvus operations or s3 storage?
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
I just browsed the source code, this warn log is printed in the milvus garbage collection loop. GC action is triggered with an interval of 1 hour. GC action recursively checks all objects under the insert_log path to see if there are segments that can be cleaned up. Seems it time out in the s3 client ListObjects() interface. By default, s3 client has a default setup with default responseHeaderTimeout = 1 minute and maxKeys = 1000. With maxKeys=1000, ListObjects() will return 1000 items for each call, maybe sometimes it timeout due to network fluctuation. You can decrease this value to a small value(50 or 100): Line 159 in b03e605 listObjectsMaxKeys=0 means using the default maxKeys for ListObjects(). |
Beta Was this translation helpful? Give feedback.
I just browsed the source code, this warn log is printed in the milvus garbage collection loop. GC action is triggered with an interval of 1 hour. GC action recursively checks all objects under the insert_log path to see if there are segments that can be cleaned up.
Seems it time out in the s3 client ListObjects() interface. By default, s3 client has a default setup with default responseHeaderTimeout = 1 minute and maxKeys = 1000. With maxKeys=1000, ListObjects() will return 1000 items for each call, maybe sometimes it timeout due to network fluctuation.
You can decrease this value to a small value(50 or 100):
milvus/configs/milvus.yaml
Line 159 in b03e605