understanding on upsert #5923
Replies: 2 comments
-
Point counts are approximate, that's explained here. Based on this the above behavior can be expected, and I don't see any clear issues. Note that the count you're seeing does not mean that there's an additional 624 points a user can retrieve. From the logical data perspective you can only see the last version of each point you've upserted. Over time, once Qdrant has done another optimization run, the count will settle at 1 million again. Also, please do not tag users directly there. They're not strictly related to this problem. |
Beta Was this translation helpful? Give feedback.
-
sorry and thanks for the reply |
Beta Was this translation helpful? Give feedback.
-
i have 1m records already indexed using hnsw
i updated few records(1k) with same id and upserted
when i seen describe collection below are the results
vectors_count=None, indexed_vectors_count=1000624, points_count=1000000
i used upsert command
so initially 1m records are there
i updated 1k records then ideally point count should be 1m (that is there) . then what about this value indexed_vectors_count=1000624,
what is wrong in it
@dzhao @bazhenov @yarikoptic @elbart @timonv
Beta Was this translation helpful? Give feedback.
All reactions