[Bug]: Insert data with duplicate primary key about milvus HOT 5 CLOSED

Milvus doesn't guarantee primary key dedup.
If both key is retrieved, new will overwrite old. but we don't guarantee this over write happens.
so user has to ensure there is no over write happens

from milvus.

Milvus doesn't guarantee primary key dedup. If both key is retrieved, new will overwrite old. but we don't guarantee this over write happens. so user has to ensure there is no over write happens

Whether or not it overwrites the old data, I think it's still within a logically controllable range. However, if a mistake is made by inserting the same primary key, as in my previous issue, when I query again using ids=[0,1], it only returns the data for id==0. Yet, the data for id==1 actually exists in the database (as it can be retrieved using ids=[1]). This seems to make the database's behavior strange and uncontrollable.

from milvus.

Milvus doesn't guarantee primary key dedup. If both key is retrieved, new will overwrite old. but we don't guarantee this over write happens. so user has to ensure there is no over write happens

Whether or not it overwrites the old data, I think it's still within a logically controllable range. However, if a mistake is made by inserting the same primary key, as in my previous issue, when I query again using ids=[0,1], it only returns the data for id==0. Yet, the data for id==1 actually exists in the database (as it can be retrieved using ids=[1]). This seems to make the database's behavior strange and uncontrollable.

because the data can be in different segment and milvus don't have the information there is a duplicate pk. this is by design and so far we can not do anything.

if you know the pk exist, you can use upsert to delete old one and insert a new one

from milvus.

Milvus doesn't guarantee primary key dedup. If both key is retrieved, new will overwrite old. but we don't guarantee this over write happens. so user has to ensure there is no over write happens

Whether or not it overwrites the old data, I think it's still within a logically controllable range. However, if a mistake is made by inserting the same primary key, as in my previous issue, when I query again using ids=[0,1], it only returns the data for id==0. Yet, the data for id==1 actually exists in the database (as it can be retrieved using ids=[1]). This seems to make the database's behavior strange and uncontrollable.

because the data can be in different segment and milvus don't have the information there is a duplicate pk. this is by design and so far we can not do anything.

if you know the pk exist, you can use upsert to delete old one and insert a new one

OK ，Thanks

from milvus.

I'd close this issue, and please free to file a new one if you met any new issues

from milvus.