[Bug]: index_null_offset of TextMatchIndex using wrong prefix and could be GCed unintentionally

[congqixia]

Is there an existing issue for this?

• I have searched the existing issues

Environment

- Milvus version: master & 2.5.4
- Deployment mode(standalone or cluster):
- MQ type(rocksmq, pulsar or kafka):    
- SDK version(e.g. pymilvus v2.0.0rc2):
- OS(Ubuntu or CentOS): 
- CPU/Memory: 
- GPU: 
- Others:

Current Behavior

The TextMatch stats logs are stored with prefix basePath/text_log. However if varchar column is nullable, an extra index file is store with basePath/index_files path which could be GCed unintentionally

Expected Behavior

index files belongs to text match index shall be stored under correct prefix

Steps To Reproduce

Milvus Log

No response

Anything else?

Related to #39889

[dosubot]

@congqixia Hello! I'm here to assist you with any bugs, questions, or contributions you have. Let's work together to resolve any issues you're experiencing.

The issue you're encountering seems to be related to the storage path configuration for index files in the TextMatchIndex. Specifically, when dealing with nullable varchar columns, the index files are being stored under the basePath/index_files path instead of the expected prefix, which could lead to them being unintentionally garbage collected.

To address this, ensure that the index files for the TextMatchIndex are stored under the correct prefix. The correct prefix for storing index files related to the TextMatchIndex is TMP_TEXT_LOG_PREFIX, defined as "/tmp/milvus/text-log/" ^[1]. Adjusting the storage path to use this prefix should prevent the files from being mistakenly garbage collected.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}