[Bug]: v2.4.0 query node auto_balance no available about milvus HOT 9 OPEN

Is there an existing issue for this?

• I have searched the existing issues

Environment

- Milvus version:2.4.0
- Deployment mode(standalone or cluster): cluster
- MQ type(rocksmq, pulsar or kafka):  kafka   
- SDK version(e.g. pymilvus v2.0.0rc2): 2.7
- OS(Ubuntu or CentOS): CentOS
- CPU/Memory: 
- GPU:  0
- Others:

Current Behavior

My Collection loading scale of 40 million entites to the mem index. I has enabled 40 query nodes, but the data is only loaded into 2 query nodes. Here is my configuration, how to adjust it to enable automatic memory balancing allocation:

Expected Behavior

No response

Steps To Reproduce

No response

Milvus Log

No response

Anything else?

No response

so you are saying you got 40m data but they got only 2 nodes with data?

1. could you offer querycoord and querynode logs?
2. could you share your segment distribution?

from milvus.

/assign @sunby
please help to follow

from milvus.

the attu-client show the error msg follow:

show collection failed: load segment failed, OOM if load, maxSegmentSize = 205.4959774017334 MB, memUsage = 121813.59168624878 MB, predictMemUsage = 122019.08766365051 MB, totalMem = 122880 MB thresholdFactor = 0.900000

from milvus.

segment:
assignmentExpiration: 2000
compactableProportion: 0.85
diskSegmentMaxSize: 2048
enableLevelZero: true
expansionRate: 1.25
maxBinlogFileNumber: 32
maxIdleTime: 600
maxLife: 86400
maxSize: 1024
minSizeFromIdleToSealed: 16
sealProportion: 0.12
smallProportion: 0.5

from milvus.

@yesyue Could you please refer this doc to export the whole Milvus logs for investigation?
Also Could you please attach the etcd backup for investigation？ Check this: https://github.com/milvus-io/birdwatcher for details about how to backup etcd with birdwatcher
/assign @yesyue
/unassign

from milvus.

the attu-client show the error msg follow:

show collection failed: load segment failed, OOM if load, maxSegmentSize = 205.4959774017334 MB, memUsage = 121813.59168624878 MB, predictMemUsage = 122019.08766365051 MB, totalMem = 122880 MB thresholdFactor = 0.900000

it seems that you can not load, might be for the unbalance reason.

you can use birdwatcher to check if all the segment is sealed or indexed.
We can not help with detailed logs and info from birdwatcher.

Using birdwatcher with show segment command can help you to figure out why

from milvus.

only one query node mem high, and increasing

querynode (3).log

from milvus.

only one query node mem high, and increasing

querynode (3).log

Can you provide querynode Segment Loaded Num and Queryable Entity Num metrics?

from milvus.

you can use birdwatcher and run download global-distribution command and paste generated distribution file here.

from milvus.