milvus
milvus copied to clipboard
[Bug]: [benchmark][standalone]Milvus insert delete scene reported error"force to deny/milvus.proto.milvus.MilvusService/Delete."
Is there an existing issue for this?
- [X] I have searched the existing issues
Environment
- Milvus version:2.2.0-20221108-2e78872b
- Deployment mode(standalone or cluster):standalone
- SDK version(e.g. pymilvus v2.0.0rc2):2.2.0dev70
- OS(Ubuntu or CentOS):
- CPU/Memory:
- GPU:
- Others:
Current Behavior
server-instance fouram-mrc8s-1 server-configmap server-single-8c32m256d-diskann client-configmap client-random-locust-compaction-high-search-diskann image:2.2.0-20221108-2e78872b pymilvus 2.2.0dev70
server:
fouram-mrc8s-1-etcd-0 1/1 Running 0 140m 10.104.6.113 4am-node13 <none> <none>
fouram-mrc8s-1-milvus-standalone-5fb6dd566f-zps2h 1/1 Running 1 (64m ago) 140m 10.104.9.142 4am-node14 <none> <none>
fouram-mrc8s-1-minio-6bcd7b6c65-xxklq 1/1 Running 0 140m 10.104.6.114 4am-node13 <none> <none>
client pod:fouram-mrc8s-369167784 client log:
[2022-11-08 13:41:44,534] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.947610', 'RPC error': '2022-11-08 13:41:44.534820'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,534] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.948445', 'RPC error': '2022-11-08 13:41:44.534975'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,535] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.948073', 'RPC error': '2022-11-08 13:41:44.535134'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,535] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.948800', 'RPC error': '2022-11-08 13:41:44.535514'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,535] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.949162', 'RPC error': '2022-11-08 13:41:44.535754'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,536] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.949489', 'RPC error': '2022-11-08 13:41:44.535994'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,536] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.949835', 'RPC error': '2022-11-08 13:41:44.536171'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,536] [ DEBUG] - Milvus get_info run in 1.1865s (milvus_benchmark.client:57)
[2022-11-08 13:41:44,536] [ DEBUG] - [scene_insert_delete_flush] Start insert sift_10w_128_l2, len:1 (milvus_benchmark.client:675)
[2022-11-08 13:41:44,536] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.950298', 'RPC error': '2022-11-08 13:41:44.536667'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,536] [ ERROR] - RPC error: [delete], <MilvusException: (code=48, message=force to deny /milvus.proto.milvus.M
ilvusService/Delete.)>, <Time:{'RPC start': '2022-11-08 13:41:43.950581', 'RPC error': '2022-11-08 13:41:44.536830'}> (pymilvus.deco
rators:108)
[2022-11-08 13:41:44,537] [ DEBUG] - Milvus get_info run in 1.1875s (milvus_benchmark.client:57)
[2022-11-08 13:41:44,537] [ DEBUG] - [scene_insert_delete_flush] Start insert sift_10w_128_l2, len:1 (milvus_benchmark.client:675)
search failed :
scene_insert_delete_flush sends about 22w requests and our server memory is given 32GB, which is more than enough memory:
Expected Behavior
No response
Steps To Reproduce
1.create an collection
2.build diskann index
3.insert 10w data
4.build index again
6.load data
7.search ,query, load, scene_insert_delete_flush->raise error
Milvus Log
No response
Anything else?
data:
config.yaml: |
locust_random_performance:
collections:
-
collection_name: sift_10w_128_l2
ni_per: 50000
other_fields: float1
build_index: true
index_type: diskann
index_param: {}
task:
types:
-
type: query
weight: 1
params:
top_k: 10
nq: 10
search_param:
search_list: 30
filters:
-
range: "{'range': {'float1': {'GT': -1.0, 'LT': collection_size * 0.5}}}"
-
type: load
weight: 1
-
type: get
weight: 1
params:
ids_length: 10
-
type: scene_insert_delete_flush
weight: 20
params:
insert_len: 1
del_len: 1
connection_num: 1
clients_num: 100
spawn_rate: 2
during_time: 72h
@time_wrapper
def scene_insert_delete_flush(self, collection_name=None, vectors=None, ids=None, del_len=30000):
if collection_name is None:
collection_name = self._collection_name
logger.debug("[scene_insert_delete_flush] Start scene_insert_delete_flush : %s" % collection_name)
collection_info = self.get_info(collection_name)
entities = utils.generate_entities(collection_info, vectors, ids)
logger.debug("[scene_insert_delete_flush] Start insert {0}, len:{1}".format(collection_name, len(ids)))
self.insert(entities, collection_name=collection_name)
logger.debug("[scene_insert_delete_flush] Start delete {0}, len:{1}".format(collection_name, del_len))
self.delete(expr="id in {}".format(ids[:del_len]), collection_name=collection_name, timeout=300)
logger.debug("[scene_insert_delete_flush] Start flush : %s" % collection_name)
self.flush(collection_name=collection_name)
logger.debug("[scene_insert_delete_flush] Finished scene_insert_delete_flush : %s" % collection_name)
@jingkl does this issue reproduce ?
This problem still exists server-instance fouram-nhkrp-1 server-configmap server-single-8c32m256d-diskann client-configmap client-random-locust-compaction-high-search-diskann
image:2.2.2-20221219-ae5259ca pymilvus :2.2.1dev4
fouram-nhkrp-1-etcd-0 1/1 Running 0 18m 10.104.5.113 4am-node12 <none> <none>
fouram-nhkrp-1-milvus-standalone-85dfbb4464-cpnlk 1/1 Running 2 (3m44s ago) 18m 10.104.4.51 4am-node11 <none> <none>
fouram-nhkrp-1-minio-69c4dc7c85-tn47k 1/1 Running 0 18m 10.104.4.52 4am-node11 <none> <none>
client log:
This problem should be associated with diskann load:#20358
/assign MrPresent-Han I will handle this
/unassign
This issue still exists. And OOM when milvus gave 8c64g
argo task: fouram-tag-no-clean-lpxx7
server-instance fouram-tag-no-clean-lpxx7-1 server-configmap server-single-8c64m256d-diskann client-configmap client-random-locust-compaction-high-search-diskann
image: 2.2.0-20230213-9c306b0d
server
fouram-tag-no-clean-lpxx7-1-etcd-0 1/1 Running 0 19h 10.104.4.45 4am-node11 <none> <none>
fouram-tag-no-clean-lpxx7-1-milvus-standalone-678bbbdb8b-rrx5c 1/1 Running 11 (18h ago) 19h 10.104.1.108 4am-node10 <none> <none>
fouram-tag-no-clean-lpxx7-1-minio-84bf878bcc-qjk85 1/1 Running 0 19h 10.104.1.107 4am-node10 <none> <none>
client pod:
fouram-tag-no-clean-lpxx7-3323611136
client log
milvus log
Anything else?
client-random-locust-compaction-high-search-diskann
collections:
-
collection_name: sift_10w_128_l2
ni_per: 50000
# other_fields: int1,int2,float1,double1
other_fields: float1
build_index: true
index_type: diskann
index_param: {}
task:
types:
-
type: query
weight: 1
params:
top_k: 10
nq: 10
search_param:
search_list: 30
filters:
-
range: \"{'range': {'float1': {'GT': -1.0, 'LT': collection_size * 0.5}}}\"
-
type: load
weight: 1
-
type: get
weight: 1
params:
ids_length: 10
-
type: scene_insert_delete_flush
weight: 20
params:
insert_len: 1
del_len: 1
connection_num: 1
clients_num: 100
spawn_rate: 2
# during_time: 84h
during_time: 12h
@xige-16 @MrPresent-Han Recorded yesterday, can you look at this question
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
.
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
.
this issue may be related to #23621 and probably has been fixed. Does the same problem reproduce now?
Verified, no recent issues with change. Verified version: 2.2.0-20230804-13f69bfc