- 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:N
- 硬件信息
- 磁盘 :机械硬盘
- CPU 8c
- 内存信息 64g
- 问题的具体描述
现象:
使用ngbatis插入数据时,出现raft buffer is full异常。
客户端有10个节点,nebula节点有11个,创建的space分片数为10个。
根据论坛相关buffer is full问题帖子参考,修改过–rocksdb_db_options={“max_subcompactions”:“4”,“max_background_jobs”:“4”} 和max_write_buff_number=8,均无法解决。具体报错日志及cpu使用截图见下.
Caused by: org.nebula.contrib.ngbatis.exception.QueryException: execute failed: ResultSet error: Storage Error: Part 4 raft buffer is full. Please retry later.
at org.nebula.contrib.ngbatis.proxy,MapperProxy.executeBySessionPool(MapperProxy.Java:297)
at org.nebula.contrib.ngbatis.proxy. MapperProxy.invoke(MapperProxy. java:134)
at org.nebula.contrib.ngbatis.proxy.MapperProxy.invoke (MapperProxy.java:170)
at org.nebula.contrib.ngbatis.proxy.MapperProxy.pageSupport (MapperProxy. java:186)
at org.nebula.contrib.ngbatis.proxy.MapperProxy.invoke(MapperProxy.java:90)
at com.pat.abc.graph.dao.CommonGraphDao$Proxy.insertEdgeIfNotExistBatch (Unknown Source)
at com.pat.abc.graph.service.CalloutGraphServiceImp1.lambda$nul1$6(CallOutGraphServiceImpl. java:287)
at com.pat.abc.component.asyncmanager.threadhelper.CommonAsyncThreadPoolHelper.submit(CommonAsyncThreadPoolHelper.java:70)
Caused by: org.nebula.contrib.ngbatis.exception,QueryException; ResultSet error: Storage Error: Part 4 raft buffer is full. Please retry later.
org.nebula. contrib.ngbatis.proxy.MapperProxy. executeBySessionPool (MapperProxy.Java:293)
... 85 common frames omitted
dashboard的cpu概览如下