1. 前言
这篇文章主要讲述如何在 docker 环境下搭建 doris 集群,以及一些注意事项。开始前请参考下必备技能表
技能槽 熟练度
docker 熟悉
dockerfile 熟悉
docker-compose 熟悉
linux 熟悉
ctrl + c , ctrl + v 精通
如果已经具备以上技能,并满足对应的熟练度要求,那么请继续,反之去熟悉一下技能
2. 简介
引用官网一句话
支持对海量大数据进行快速分析的 MPP 数据库
官网地址:https://doris.apache.org
3. 编译
由于官方没有直接给出编译后的程序,所以只能自己编译后使用
个人猜测应该是在 apache 还处于孵化阶段的原因。
本教程使用官方比较推荐的 docker方式编译
3.1. 源码下载
前往github下载 Apache Doris 0.14.0 (incubating) Release 源码
3.2. 环境搭建
将源码解压至当前目录的source/0.14.0下
如对 docker-compose 比较了解的话,目录放哪都无所谓,只要能挂在到容器的 /root/source 即可
创建 build.yml并键入以下内容
version: "3.7"
networks:
doris-build:
name: doris-build
driver: bridge
services:
doris-build:
image: apachedoris/doris-dev:build-env-1.2
container_name: dev
hostname: dev
privileged: true
command: ["source/0.14.0/build.sh"]
volumes:
- ./data/.m2:/root/.m2
- ./source:/root/source
networks:
- doris-build
3.3. 编译源码
执行 docker-compose -f build.yml up,等待执行完毕,整个过程大概需要十分钟左右,因此第一次需要到mavan下载第三方库。
如果期间遇到Could not transfer artifact net.sourceforge.czt.dev:cup-maven-plugin:pom:1.6-cdh from/to xxx问题,根据 github 上的 PR 修改下 fe/pom.xml 的 cloudera 仓库
此时你会在源码的目录下得到output文件夹,目录结构如下
source/0.14.0/output/
├── be
│ ├── bin
│ ├── conf
│ ├── lib
│ └── www
├── fe
│ ├── bin
│ ├── conf
│ ├── lib
│ ├── spark-dpp
│ └── webroot
└── udf
├── include
└── lib
4. 运行
鉴于编译环境用 docker,那么我们运行的时候也使用 docker
4.1. 制作镜像
新建Dockerfile并键入以下内容
FROM centos:7.5.1804
RUN \
yum clean metadata && \
yum -y install epel-release && \
yum clean metadata && \
yum -y install make which tar rpm-build yum-utils python-argparse python-yaml java-1.8.0-openjdk-devel gcc gcc-c|| createrepo jq glibc-static krb5-workstation openssh-clients && \
yum clean all
ENV JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk
RUN mkdir -p /home/doris
COPY ./source/0.14.0/output/fe /home/doris/fe
COPY ./source/0.14.0/output/be /home/doris/be
COPY ./source/0.14.0/fs_brokers/apache_hdfs_broker/output/apache_hdfs_broker /home/doris/fs_broker
运行docker build --pull --rm -f Dockerfile -t doris:0.14.00 .编译镜像
4.2. 编排容器
新建 docker-compose.yml并键入以下内容
version: "3.7"
networks:
doris:
name: doris
driver: bridge
services:
fe-leader:
image: doris:0.14.00
restart: always
container_name: fe-leader
hostname: fe-leader
environment:
- TZ=Asia/Shanghai
command: /home/doris/fe/bin/start_fe.sh
volumes:
- ./doris/fe/leader/log:/home/doris/fe/log
- ./doris/fe/leader/doris-meta:/home/doris/fe/doris-meta
- ./doris/fe/leader/conf:/home/doris/fe/conf
- /etc/localtime:/etc/localtime:ro
stdin_open: true
tty: true
ports:
- 8030:8030
- 8040:8040
- 9030:9030
networks:
- doris
fe-follower-1:
image: doris:0.14.00
restart: always
container_name: fe-follower-1
hostname: fe-follower-1
environment:
- TZ=Asia/Shanghai
command: /home/doris/fe/bin/start_fe.sh --helper fe-leader:9010
volumes:
- ./doris/fe/fe-follower-1/log:/home/doris/fe/log
- ./doris/fe/fe-follower-1/doris-meta:/home/doris/fe/doris-meta
- ./doris/fe/fe-follower-1/conf:/home/doris/fe/conf
- /etc/localtime:/etc/localtime:ro
stdin_open: true
tty: true
security_opt:
- seccomp:unconfined
privileged: true
depends_on:
- fe-leader
networks:
- doris
fe-observer-1:
image: doris:0.14.00
restart: always
container_name: fe-observer-1
hostname: fe-observer-1
environment:
- TZ=Asia/Shanghai
command: /home/doris/fe/bin/start_fe.sh --helper fe-leader:9010
volumes:
- ./doris/fe/fe-observer-1/log:/home/doris/fe/log
- ./doris/fe/fe-observer-1/doris-meta:/home/doris/fe/doris-meta
- ./doris/fe/fe-observer-1/conf:/home/doris/fe/conf
- /etc/localtime:/etc/localtime:ro
stdin_open: true
tty: true
security_opt:
- seccomp:unconfined
privileged: true
depends_on:
- fe-leader
networks:
- doris
be:
image: doris:0.14.00
restart: always
container_name: be
hostname: be
environment:
- TZ=Asia/Shanghai
command: /home/doris/be/bin/start_be.sh
volumes:
- "./doris/be/log:/home/doris/be/log"
- "./doris/be/storage:/home/doris/be/storage"
- "./doris/be/conf:/home/doris/be/conf/"
- "/etc/localtime:/etc/localtime:ro"
stdin_open: true
tty: true
security_opt:
- seccomp:unconfined
depends_on:
- fe-leader
networks:
- doris
mysql:
image: "mysql:5.7"
container_name: mysql
restart: always
environment:
MYSQL_ROOT_PASSWORD: "123456"
MYSQL_USER: "root"
MYSQL_PASS: "123456"
TZ: Asia/Shanghai
volumes:
- ./doris/mysql:/var/lib/mysql
- ./doris/mysql:/docker-entrypoint-initdb.d/
- ./doris/mysql/sources.list:/etc/apt/sources.list
depends_on:
- fe-leader
networks:
- doris
4.3. 运行容器
新建doris/fe文件夹并保持如下结构
doris/fe/
├── fe-follower-1
│ ├── conf
│ ├── doris-meta
│ └── log
├── fe-observer-1
│ ├── conf
│ ├── doris-meta
│ └── log
└── leader
├── conf
├── doris-meta
└── log
新建doris/be/conf文件夹
复制前面编译出来的 output/fe/conf/fe.conf 到对应的 fe 的 conf 文件夹里
复制output/be/conf/be.conf 到对应的 conf 文件夹里
执行docker-compose up -d
5. 配置
当容器运行后,各容器之间是独立运行的,需要将follower、oberver、be添加到leader中才能形成集群,如果只不需要集群可以只添加be至leader,添加前后顺序可以随便
leader follower oberver 相关资料请查阅官方文档,这里就不阐述了
当且必须在be被添加到leader后,可使用mysql数据库连接工具如navcat、 dbeaver连接到leader的9010端口,该端口是mysql协议,默认账号为root,密码为空
5.1. 添加 follower
执行docker exec -it mysql /bin/bash进入mysql容器
执行脚本将follower添加到leader中
mysql -h fe-leader -u root -P 9030 <<EOF
ALTER SYSTEM ADD FOLLOWER "fe-follower-1:9010";
EOF
5.2. 添加 observer
同样在mysql容器内执行脚本
mysql -h fe-leader -u root -P 9030 <<EOF
ALTER SYSTEM ADD OBSERVER "fe-observer-1:9010";
EOF
5.3. 添加 be
依旧在mysql容器内执行脚本
mysql -h fe-leader -u root -P 9030 <<EOF
ALTER SYSTEM ADD BACKEND "be:9050";
EOF
5.4. 验证各容器运行状态
在mysql容器内执行脚本查看所有 fe 节点信息
mysql -h fe-leader -u root -P 9030 <<EOF
SHOW PROC '/frontends';
EOF
输出内容如下
Name IP HostName EditLogPort HttpPort QueryPort RpcPort Role IsMaster ClusterId Join Alive ReplayedJournalId LastHeartbeat IsHelper ErrMsg Version
172.26.0.6_9010_1624243783898 172.26.0.6 fe-observer-1.doris 9010 8030 9030 9020 OBSERVER false 1667475297 true true 557 2021-06-21 11:00:55 false 0.14.0-release-Unknown
172.26.0.3_9010_1624243326588 172.26.0.3 fe-follower-1.doris 9010 8030 9030 9020 FOLLOWER false 1667475297 true true 557 2021-06-21 11:00:55 true 0.14.0-release-Unknown
172.26.0.2_9010_1624242600423 172.26.0.2 fe-leader 9010 8030 9030 9020 FOLLOWER true 1667475297 true true 558 2021-06-21 11:00:55 true 0.14.0-release-Unknown
如果Alive列为true则标识已成功添加到leader中,反之则失败,此时需要查看对应日志fe.log文件,根据具体信息解决
doris/fe/
├── fe-follower-1
│ ├── conf
│ │ ├── fe.conf
│ │ └── log4j2-spring.xml
│ ├── doris-meta
│ │ ├── bdb
│ │ └── image
│ └── log
│ ├── fe.audit.log
│ ├── fe.gc.log.20210621-103001
│ ├── fe.log
│ ├── fe.out
│ └── fe.warn.log
├── fe-observer-1
│ ├── conf
│ │ ├── fe.conf
│ │ └── log4j2-spring.xml
│ ├── doris-meta
│ │ ├── bdb
│ │ └── image
│ └── log
│ ├── fe.audit.log
│ ├── fe.gc.log.20210621-104708
│ ├── fe.gc.log.20210621-105052
│ ├── fe.log
│ ├── fe.out
│ └── fe.warn.log
└── leader
├── conf
│ ├── fe.conf
│ ├── log4j2-spring.xml
│ ├── table1_data
│ ├── table2_data
│ ├── tools
│ └── up.sh
├── doris-meta
│ ├── bdb
│ └── image
└── log
├── fe.audit.log
├── fe.gc.log.20210621-102958
├── fe.log
├── fe.out
└── fe.warn.log
在mysql容器内执行脚本查看所有 be 节点信息
mysql -h fe-leader -u root -P 9030 <<EOF
SHOW PROC '/backends';
EOF
输出内容如下
BackendId Cluster IP HostName HeartbeatPort BePort HttpPort BrpcPort LastStartTime LastHeartbeat Alive SystemDecommissioned ClusterDecommissioned TabletNum DataUsedCapacity AvailCapacity TotalCapacity UsedPct MaxDiskUsedPct ErrMsg Version Status
10004 default_cluster 172.26.0.4 be.doris 9050 9060 8040 8060 2021-06-21 10:40:42 2021-06-21 16:09:01 true false false 0 .000 140.327 GB 263.000 GB 46.64 % 46.64 % 0.14.0-release-Unknown {“lastSuccessReportTabletsTime”:“2021-06-21 16:08:22”}
结语
不出意外,到这里你已经完成了 doris 的集群搭建了,至此本教程的使命也完成了。关于如何使用还需要查阅官方文档。
常见问题
1. wait catalog to be ready
fe启动后日志不断输出如下内容
2021-06-24 14:14:56,907 INFO (UNKNOWN 172.26.0.6_9010_1624243783898(-1)|1) [Catalog.waitForReady():785] wait catalog to be ready. FE type: UNKNOWN. is ready: false
2021-06-24 14:14:58,908 INFO (UNKNOWN 172.26.0.6_9010_1624243783898(-1)|1) [Catalog.waitForReady():785] wait catalog to be ready. FE type: UNKNOWN. is ready: false
2021-06-24 14:15:00,909 INFO (UNKNOWN 172.26.0.6_9010_1624243783898(-1)|1) [Catalog.waitForReady():785] wait catalog to be ready. FE type: UNKNOWN. is ready: false
2021-06-24 14:15:02,910 INFO (UNKNOWN 172.26.0.6_9010_1624243783898(-1)|1) [Catalog.waitForReady():785] wait catalog to be ready. FE type: UNKNOWN. is ready: false
2021-06-24 14:15:04,911 INFO (UNKNOWN 172.26.0.6_9010_1624243783898(-1)|1) [Catalog.waitForReady():785] wait catalog to be ready. FE type: UNKNOWN. is ready: false
解决方法
或者参考 github#5292 的解决方法
清除非Master节点的元数据
在Master节点的FE的配置文件fe.conf文件添加如下内容并重启所有FE
metadata_failure_recovery = true
等待各节点都正常后启动并通过SHOW PROC '/frontends';确保每个节点的Join和Alive都为true后,注释掉第二步的设置,避免出现不必要的问题
2. Key columns should be a ordered prefix of the schema
建表时输出如下异常
Key columns should be a ordered prefix of the schema
建表语句如下
CREATE TABLE Location
(
CREATE TABLE `Location` (
`Id` varchar(36) NOT NULL COMMENT "主键",
`Latitude` int(11) NOT NULL COMMENT "纬度信息,以度为单位的纬度值乘以 10 的 6 次方,精确到百万分之一度",
`Longitude` int(11) NOT NULL COMMENT "经度信息,以度为单位的经度值乘以 10 的 6 次方,精确到百万分之一度",
`Altitude` smallint(6) NOT NULL COMMENT "高程(海拔高度),海拔高度,单位为米(m)",
`Speed` smallint(6) NOT NULL COMMENT "速度,1/10km/h",
`Direction` smallint(6) NOT NULL COMMENT "方向,0-359,正北为 0,顺时针",
`DateTime` datetime NOT NULL COMMENT "终端上报时间",
`SimNumber` varchar(15) NULL COMMENT "sim卡号",
`IsBatch` boolean NOT NULL COMMENT "是否为补传消息",
`CreateDateTime` datetime NOT NULL COMMENT "创建时间"
) ENGINE=OLAP
UNIQUE KEY(`Id`, `SimNumber`)
COMMENT "OLAP"
DISTRIBUTED BY HASH(`SimNumber`) BUCKETS 10
PROPERTIES (
"replication_num" = "1",
"in_memory" = "false",
"storage_format" = "V2"
);
解决方法
只需要修改语句中列的顺序即可
将KEY中的属性移至到前面,如本例中将SimNumber` varchar(15) NULL COMMENT "sim卡号",`移至Idvarchar(36) NOT NULL COMMENT "主键",后面即可
CREATE TABLE `Location` (
`Id` varchar(36) NOT NULL COMMENT "主键",
`SimNumber` varchar(15) NULL COMMENT "sim卡号",
`Latitude` int(11) NOT NULL COMMENT "纬度信息,以度为单位的纬度值乘以 10 的 6 次方,精确到百万分之一度",
`Longitude` int(11) NOT NULL COMMENT "经度信息,以度为单位的经度值乘以 10 的 6 次方,精确到百万分之一度",
`Altitude` smallint(6) NOT NULL COMMENT "高程(海拔高度),海拔高度,单位为米(m)",
`Speed` smallint(6) NOT NULL COMMENT "速度,1/10km/h",
`Direction` smallint(6) NOT NULL COMMENT "方向,0-359,正北为 0,顺时针",
`DateTime` datetime NOT NULL COMMENT "终端上报时间",
`IsBatch` boolean NOT NULL COMMENT "是否为补传消息",
`CreateDateTime` datetime NOT NULL COMMENT "创建时间"
) ENGINE=OLAP
UNIQUE KEY(`Id`, `SimNumber`)
COMMENT "OLAP"
DISTRIBUTED BY HASH(`SimNumber`) BUCKETS 10
PROPERTIES (
"replication_num" = "1",
"in_memory" = "false",
"storage_format" = "V2"
);
注意事项
容器最好使用静态IP方式,因为这玩意比较依赖IP
使用Stream load方式导入数据时请确保数据中没有包含',否则会报value is null while columns is not nullable错误