提问参考模版:
- nebula 版本:3.4.1
- 部署方式:分布式
- 安装方式:RPM
- 是否上生产环境:Y
- 硬件信息
- 磁盘( 推荐使用 SSD)
- CPU、内存信息
- 问题的具体描述
目前Space 上同一个点有两种tag, Group 和 Compacted,由于vid 相同,我用get subgraph 语句的时候发现会同时获取Group 和 Compacted 两个实体的属性,这个导致性能很差
,请问get subgraph 目前有什么办法可以过滤不出现Group 这个tag 类型吗?
提问参考模版:
性能很差
,请问get subgraph 目前有什么办法可以过滤不出现Group 这个tag 类型吗?如果考虑性能的话,分开两步?
第一步查询不带 with prop。然后第二步按需取 prop
如果是为了可视化查询感觉这种方法是可行的。可以先在画布上做呈现,然后鼠标移动上去显示 prop。
不太可行,因为拿到数据之后要根据prop 做业务逻辑处理
profile看看?
"id","name","dependencies","profiling data","operator info"
"2","DataCollect","1","ver: 0, rows: 11, execTime: 1322399us, totalTime: 1322416us","outputVar: {
""colNames"": [
""nodes"",
""rels""
],
""type"": ""DATASET"",
""name"": ""__DataCollect_2""
}
inputVar: [
{
""colNames"": [],
""type"": ""DATASET"",
""name"": ""__Subgraph_1""
}
]
distinct: false
kind: SUBGRAPH"
"1","Subgraph","0","{
ver: 0, rows: 17452, execTime: 298us, totalTime: 2702654us
total_rpc_time: 11215(us)
resp[0]: {
""exec"": ""1087(us)"",
""host"": ""172.18.103.120:9779"",
""storage_detail"": {
""GetNeighborsNode"": ""690(us)"",
""HashJoinNode"": ""260(us)"",
""RelNode"": ""691(us)"",
""SingleEdgeNode"": ""88(us)"",
""TagNode"": ""157(us)""
},
""total"": ""10827(us)"",
""vertices"": 1
}
resp[2]: {
""exec"": ""3429(us)"",
""host"": ""172.18.103.120:9779"",
""storage_detail"": {
""GetNeighborsNode"": ""7027(us)"",
""HashJoinNode"": ""5760(us)"",
""RelNode"": ""7033(us)"",
""SingleEdgeNode"": ""2836(us)"",
""TagNode"": ""2367(us)""
},
""total"": ""8219(us)"",
""vertices"": 39
}
resp[1]: {
""exec"": ""1835(us)"",
""host"": ""172.18.103.125:9779"",
""storage_detail"": {
""GetNeighborsNode"": ""4959(us)"",
""HashJoinNode"": ""3766(us)"",
""RelNode"": ""4966(us)"",
""SingleEdgeNode"": ""1610(us)"",
""TagNode"": ""1821(us)""
},
""total"": ""4283(us)"",
""vertices"": 28
}
}","outputVar: {
""colNames"": [],
""type"": ""DATASET"",
""name"": ""__Subgraph_1""
}
src: COLUMN[0]
tag_filter:
edge_filter:
filter:
vertexProps: [
{
""props"": [
""name"",
""econkind"",
""isipo"",
""shareholder"",
""create_time"",
""_tag""
],
""tagId"": 70
},
{
""props"": [
""name"",
""hasimage"",
""supernode"",
""create_time"",
""_tag""
],
""tagId"": 71
}
]
edgeProps: [
{
""props"": [
""_src"",
""_type"",
""_rank"",
""_dst"",
""midid"",
""role"",
""stockpercent"",
""shouldcapi"",
""create_time""
],
""type"": 73
}
]
steps: 10"
"0","Start","","ver: 0, rows: 0, execTime: 0us, totalTime: 78us","outputVar: {
""colNames"": [],
""type"": ""DATASET"",
""name"": ""__Start_0""
}"
上面是profile 的结果,我看subgraph 算子也才耗时2s, 为什么整体耗时要8s左右?
意外的是: 我把20层的数据,压缩到了10层,以为耗时会有所优化,但是发现并没有什么区别,这个怎么理解?
有人可以看下这个跟问题怎么分析吗
此话题已在最后回复的 30 天后被自动关闭。不再允许新回复。