[Bug]: Query iterator failed to get entire result set

[MrPresent-Han]

Is there an existing issue for this?

• I have searched the existing issues

Environment

- Milvus version:2.3.3
- Deployment mode(standalone or cluster):standalone
- MQ type(rocksmq, pulsar or kafka):    
- SDK version(e.g. pymilvus v2.0.0rc2):
- OS(Ubuntu or CentOS): 
- CPU/Memory: 
- GPU: 
- Others:

Current Behavior

Using query iterator and try to retrieve all matched results, but failed with only a small part of

Expected Behavior

No response

Steps To Reproduce

import pymilvus
from pymilvus import connections, Collection

print(pymilvus.__version__)

collection_name = 'prod_embed'

# Connect to the Milvus server
conn = connections.connect(uri=mappings[collection_name]['uri'], token=mappings[collection_name]['token'])

# Load the collection
zilliz_collection = Collection(collection_name)
zilliz_collection.load()

results = zilliz_collection.query(expr='unique_id != ""',
                                  output_fields=['unique_id', 'timestamp'])

# outputs 955037
print(len(results))


# Iterate through results
query_iterator = zilliz_collection.query_iterator(
        batch_size=10000,
        expr=f"unique_id != '' ",
        output_fields=['unique_id', 'timestamp'])

count = 0
# Iterate through results
while True:
    res = query_iterator.next()
    if len(res) == 0:
        print(f"Query iteration finished, close: {count}")
        query_iterator.close()
        break

    count += len(res)
    print(count)

# finally outputs only 132343
print(count)

Milvus Log

No response

Anything else?

No response

[yanliang567]

/assign @MrPresent-Han

/assign @NicoYuan1986
@NicoYuan1986 please help to check whether we need a new test for this scenario?

[NicoYuan1986]

strange.
milvus: master-20231226-033456ea
pymilvus: 2.4.0rc9

>>> expression = "pk != -1"
>>> query_iterator = collection.query_iterator(expr=expression, batch_size=1000)
>>> 
>>> page_idx = 0
>>> while True:
...     res = query_iterator.next()  # next翻页
...     if len(res) == 0:
...         print("query iteration finished, close")
...         query_iterator.close()  # 没有新的结果，翻页结束，close掉iterator
...         break
...     page_idx += 1
...     print(f"page{page_idx}-------------------------{len(res)}")
... 
page1-------------------------500
page2-------------------------500
page3-------------------------334
page4-------------------------334
page5-------------------------334
page6-------------------------334
page7-------------------------334
page8-------------------------334
page9-------------------------334
page10-------------------------334
page11-------------------------334
page12-------------------------334
page13-------------------------334
page14-------------------------348
page15-------------------------978
query iteration finished, close

[yanliang567]

Verified not repro on 2.3-20231227-eb11b1a5

1. insert 1m data with pk(1-1m) and wait a few hours to make sure all segments compacted done
2. insert 1m data with the same PK(1-1m), wait a few hours to make sure all compaction done
3. check num_enetities = 2m
4. check query(count(*))=2m
5. query iterator returns 200 pages with 5000 entities each page, which means the total entities(after dedup) was retrieved successfully.