1、dse修改内存到16G,修改kjdse.sh
-Xmx16G
-Xms16G
-Xmn5G
-XX:PermSize=256M
-XX:MaxPermSize=512M
2 网关修改配置文件 config.properties 修改
## processor\u6570\u91cf
processorNum=4
threadNum=28
openBusiFlag=1
sendThreadNum=1
修改 log4j.properties
log4j.rootLogger=error,cpsdna
log4j.appender.cpsdna.BufferSize=8192
GW的tomcat下/app/gw/bin/catalina.sh
JAVA_OPTS="-Xms1024M -Xmx3072M -Xss228K -XX:PermSize=512M -XX:MaxPermSize=2048M"
CATALINA_OPTS='-Djava.nio.channels.spi.SelectorProvider=sun.nio.ch.EPollSelectorProvider
' 修改epoll性能
3.gw 增加虚拟内存
4.ulimit -n 655350 GW
5,kafka,dse 设为warn
6.kafka 删除数据策略
log.cleanup.policy=delete
log.retention.hours=6
7.dse在未解析地址时配置
baidu_geo_to_addr_url=http://api.map.baidu.com/geocoder/v2/?location=%s,%s&output=json&ak=kGV9zPXCeAcOcmbxR89U17F8bGi9nou0
8.dse优化数据库,加大批量大小
9. 数据库5.7也要重新安装 tokudb配置
10加了个字段,加了个索引 (车次统计)
ALTER TABLE study_base.kajia_vehicle_entertimes add TYPE int(2) DEFAULT 0 COMMENT '类型:0:跨省,1:跨市' after VEHICLE_ID ;
update study_base.kajia_vehicle_entertimes set type = 1 where city_name is not null;
ALTER TABLE study_base.kajia_vehicle_entertimes add INDEX index_vehicleId_startTime(VEHICLE_ID,TYPE);