[Bug]: [benchmark][standalone] Enabled eviction & all mmap, during serial search, milvus disk usage exceeds the limit 10Gi, causing pod to be evicted #43123
Description
Is there an existing issue for this?
- I have searched the existing issues
Environment
- Milvus version:master-20250701-26ec841f-amd64
- Deployment mode(standalone or cluster):standalone
- MQ type(rocksmq, pulsar or kafka):rocksmq
- SDK version(e.g. pymilvus v2.0.0rc2):2.6.0rc151
- OS(Ubuntu or CentOS):
- CPU/Memory:
- GPU:
- Others:Current Behavior
argo task: fouramf-fj2gd
server:
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
verify-cl-evict-2-etcd-0 1/1 Running 0 2d3h 10.104.20.239 4am-node22 <none> <none>
verify-cl-evict-2-milvus-standalone-7fd777c48c-54m5j 0/1 Error 0 2d3h 10.104.23.133 4am-node27 <none> <none>
verify-cl-evict-2-milvus-standalone-7fd777c48c-x4kcx 1/1 Running 0 23h 10.104.23.85 4am-node27 <none> <none>
verify-cl-evict-2-minio-745fccf4d4-4cbbn 1/1 Running 0 2d3h 10.104.20.240 4am-node22 <none> <none>
Expected Behavior
No response
Steps To Reproduce
1. create a collection with fields: 'id'(primary key), 'float_vector'(128dim)
2. insert 50 million data
3. flush collection
4. build HNSW index on vector field
5. load collection
6. serial search with different params -> milvus disk usage exceeds the limit 10GiMilvus Log
No response
Anything else?
server config:
cluster:
enabled: false
etcd:
metrics:
enabled: true
podMonitor:
enabled: true
replicaCount: 1
extraConfigFiles:
user.yaml: |-
queryNode:
segcore:
tieredStorage:
evictionEnabled: true
mmap:
scalarField: true
scalarIndex: true
vectorField: true
vectorIndex: true
dataCoord:
segment:
sealProportion: 1
indexCoord:
scheduler:
interval: 1
image:
all:
repository: harbor.milvus.io/milvus/milvus
tag: master-20250701-26ec841f-amd64
log:
level: debug
metrics:
serviceMonitor:
enabled: true
minio:
metrics:
podMonitor:
enabled: true
mode: standalone
pulsarv3:
enabled: false
standalone:
disk:
size:
enabled: true
resources:
limits:
cpu: "4.0"
ephemeral-storage: 10Gi
memory: 8Gi
requests:
cpu: "2.0"
memory: 8Gi
streaming:
enabled: true
client config:
{
"dataset_params": {
"metric_type": "L2",
"dim": 128,
"dataset_name": "sift",
"dataset_size": "50m",
"ni_per": 50000,
"req_run_counts": 10
},
"collection_params": {
"other_fields": [],
"shards_num": 2
},
"search_params": {
"timeout": 36000,
"top_k": [
10,
100
],
"nq": 10000,
"search_param": {
"ef": [
128,
256
]
}
},
"index_params": {
"index_type": "HNSW",
"index_param": {
"M": 16,
"efConstruction": 500
}
}
}