[Bug]: newly inserted data does not display the correct segment row numbers

[douglarek]

Is there an existing issue for this?

• I have searched the existing issues

Environment

- Milvus version: 2.3.13
- Deployment mode(standalone or cluster): cluster
- MQ type(rocksmq, pulsar or kafka):    kafka
- SDK version(e.g. pymilvus v2.0.0rc2):
- OS(Ubuntu or CentOS): 
- CPU/Memory: enough
- GPU: 
- Others:

Current Behavior

Executing flush on attu, after writing to disk, shows no data.

But it also shows that the actual number of inserted entries is 445,000.

And what's even more absurd is that although it clearly states 1-22 data entries, only 16 are actually displayed.

Expected Behavior

No response

Steps To Reproduce

No response

Milvus Log

No response

Anything else?

I'm not sure if this is an issue with attu（version 2.3.10） or not. For attu 2.3.8 and 2.3.10, the segment section always displays 0 rows.

[yanliang567]

/assign @shanghaikid
please help to check if it is a attu issue.
/unassign

[xiaofan-luan]

@douglarek I'm assuming this is more like a bug on your usage.
Could you share you code about how you implement?

[douglarek]

Could you share you code about how you implement?

How do I understand the code you mentioned sharing, in terms of which aspect of the code?

By the way, I have a collection of 500,000 vectors. Reloading and releasing them is not an issue, but the segmentation display is problematic.

Is it normal that a collection of 500,000 vectors doesn't generate any segments?

And next, how should I go about debugging to identify where the issue lies?

[douglarek]

Update: It is now essentially confirmed that the issue is likely not related to Attu. After killing the querynod with excessively high memory usage, the displayed count of loaded collections directly shows 0, making retrieval impossible. I'm curious if we've ever conducted high-dimensional collection tests at a large data scale.

[xiaofan-luan]

@douglarek
you have to do a manual flush or wait for 24 hours or fill the size of one segment before you can see it.
But that don't affect you search.
I you want to know the exact number of collection, use count.

[douglarek]

@douglarek you have to do a manual flush or wait for 24 hours or fill the size of one segment before you can see it. But that don't affect you search. I you want to know the exact number of collection, use count.

Why do I have to manually flush? Since Mivus 2.x, the official recommendation is not to manually flush, and what is the mechanism for waiting 24 hours?

[xiaofan-luan]

if you want to check the correct num entities. then flush could help.
better way is to use count

[douglarek]

if you want to check the correct num entities. then flush could help. better way is to use count

Please take a look at the first image I posted in this issue. Even though I flushed it and it shows how many segments there are, each segment has zero rows, which is unsettling.

[xiaofan-luan]

what is the number you get if you use pymilvus client?

if it's also 0, could you try to use birdwatcher and check the how entity is there i your meta of your segment?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Rotten issues close after 30d of inactivity. Reopen the issue with /reopen.