From 419868d0edb17a9f7724a79e85ced745430fb783 Mon Sep 17 00:00:00 2001 From: lijun Date: Mon, 8 Mar 2021 10:49:45 +0800 Subject: [PATCH] add documents --- ...06\345\214\272\346\200\247\350\203\275.md" | 232 +++++++++++++ .../post/jiajunfeng/openGauss+KeepAlived.md | 304 ++++++++++++++++++ 2 files changed, 536 insertions(+) create mode 100644 "content/zh/post/gaoyunlong/PostgreSQL\344\270\216openGauss\344\271\213\345\210\206\345\214\272\346\200\247\350\203\275.md" create mode 100644 content/zh/post/jiajunfeng/openGauss+KeepAlived.md diff --git "a/content/zh/post/gaoyunlong/PostgreSQL\344\270\216openGauss\344\271\213\345\210\206\345\214\272\346\200\247\350\203\275.md" "b/content/zh/post/gaoyunlong/PostgreSQL\344\270\216openGauss\344\271\213\345\210\206\345\214\272\346\200\247\350\203\275.md" new file mode 100644 index 00000000..fc62d44f --- /dev/null +++ "b/content/zh/post/gaoyunlong/PostgreSQL\344\270\216openGauss\344\271\213\345\210\206\345\214\272\346\200\247\350\203\275.md" @@ -0,0 +1,232 @@ ++++ + +title = "PostgreSQL与openGauss之分区性能" + +date = "2021-03-08" + +tags = ["openGauss与PostgreSQL对比"] + +archives = "2021-03" + +author = "高云龙" + +summary = "PostgreSQL与openGauss之分区性能" + +img = "/zh/post/gaoyunlong/title/img25.png" + +times = "10:40" + ++++ + +# PostgreSQL与openGauss之分区性能 + +## 概述 + +PostgreSQL与openGauss分区表定义差异，请参考https://www.modb.pro/db/41393。 + +openGauss1.1.0开始支持hash/list分区，hash分区表最多支持64个分区，否则会报： + +``` +ERROR: Un-support feature +DETAIL: The partition’s length should be less than 65. +``` + +本次对PostgreSQL 和 openGauss 64个子分区表的常规操作对比。 + +服务器配置：虚拟机8G4C50G + +数据库版本：PostgreSQL13.1、openGauss1.1.0 + +## 添加分区表 + +PostgreSQL数据库： + +``` +--创建父表 +CREATE TABLE partition_table( + id int, + col1 character varying(16), + create_time timestamptz +) PARTITION BY HASH(id); + +--添加分区 +SELECT 'CREATE TABLE partition_table_' || n || ' PARTITION of partition_table FOR VALUES WITH (MODULUS 64, REMAINDER ' || n || ');' FROM generate_series(0,63) as n ;\gexec + +--初始化数据 +INSERT INTO partition_table(id,col1,create_time) SELECT round(100000000*random()), n || '_col1',now() FROM generate_series(1,10000000) n; + +--添加索引 +CREATE INDEX ON partition_table USING BTREE(id); +CREATE INDEX ON partition_table USING BTREE(col1); +``` + +openGauss数据库： + +``` +--创建分区表 +create table partition_table( + id int, + col1 varchar(16), + create_time timestamptz default now()) +partition by hash(id) +(partition part_hash_1, +partition part_hash_2, +partition part_hash_3, +partition part_hash_4, +partition part_hash_5, +partition part_hash_6, +partition part_hash_7, +partition part_hash_8, +partition part_hash_9, +partition part_hash_10, +partition part_hash_11, +partition part_hash_12, +partition part_hash_13, +partition part_hash_14, +partition part_hash_15, +partition part_hash_16, +partition part_hash_17, +partition part_hash_18, +partition part_hash_19, +partition part_hash_20, +partition part_hash_21, +partition part_hash_22, +partition part_hash_23, +partition part_hash_24, +partition part_hash_25, +partition part_hash_26, +partition part_hash_27, +partition part_hash_28, +partition part_hash_29, +partition part_hash_30, +partition part_hash_31, +partition part_hash_32, +partition part_hash_33, +partition part_hash_34, +partition part_hash_35, +partition part_hash_36, +partition part_hash_37, +partition part_hash_38, +partition part_hash_39, +partition part_hash_40, +partition part_hash_41, +partition part_hash_42, +partition part_hash_43, +partition part_hash_44, +partition part_hash_45, +partition part_hash_46, +partition part_hash_47, +partition part_hash_48, +partition part_hash_49, +partition part_hash_50, +partition part_hash_51, +partition part_hash_52, +partition part_hash_53, +partition part_hash_54, +partition part_hash_55, +partition part_hash_56, +partition part_hash_57, +partition part_hash_58, +partition part_hash_59, +partition part_hash_60, +partition part_hash_61, +partition part_hash_62, +partition part_hash_63, +partition part_hash_64); + +--初始化数据 +INSERT INTO partition_table(id,col1,create_time) SELECT round(100000000*random()), n || '_col1',now() FROM generate_series(1,10000000) n; + +--添加全局索引 +CREATE INDEX ON partition_table USING BTREE(id); +CREATE INDEX ON partition_table USING BTREE(col1); + +--添加本地索引 +CREATE INDEX ON partition_table USING BTREE(id) local; +CREATE INDEX ON partition_table USING BTREE(col1) local; +``` + +## 测试方法 + +采用pgbench压测工具，自定义压测脚本的方式来对比。 + +``` +cat bench.sql +\set idpp random(1,100000) +--insert into partition_table values(:idpp,:idpp||'_col1',now()); +--update partition_table set create_time=now() where id=:idpp; +--update partition_table set create_time=now() where col1=:idpp||'_col1'; +--select * from partition_table where id=:idpp; +--select * from partition_table where col1=:idpp||'_col1'; + +pgbench -p 5432 -j 30 -c 30 -M prepared -T 30 -n yunlong -f bench.sql +``` + +## 结果对比 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
  

分区键查询

+

非分区键查询

+

分区键更新

+

非分区键更新

+

插入

+

PostgreSQL

+

0.594 ms

+

7.978 ms

+

1.612 ms

+

17.413 ms

+

17.2ms

+

openGauss（全局索引）

+

0.612 ms

+

0.758 ms

+

10.450 ms

+

88.151 ms

+

78.082 ms

+

openGauss（本地索引）

+

5.635 ms

+

6.765 ms

+

15.187 ms

+

94.614 ms

+

84.927 ms

+
+ +结果对比发现， + +1、Postgresql13.1 版本在分区方面总来看优越于openGauss1.1.0。 + +2、opengauss 全局索引会比本地索引性能更好，但全局索引维护成本高。 + +3、非分区键查询，带全局索引的opengauss查询性能最快。 + +此测试受限于服务器环境，数据仅做参考比对。 + diff --git a/content/zh/post/jiajunfeng/openGauss+KeepAlived.md b/content/zh/post/jiajunfeng/openGauss+KeepAlived.md new file mode 100644 index 00000000..54dee04e --- /dev/null +++ b/content/zh/post/jiajunfeng/openGauss+KeepAlived.md @@ -0,0 +1,304 @@ ++++ + +title = "openGauss+KeepAlived" + +date = "2021-03-08" + +tags = ["openGauss+KeepAlived"] + +archives = "2021-03" + +author = "贾军锋" + +summary = "openGauss+KeepAlived" + +img = "/zh/post/jiajunfeng/title/img33.jpg" + +times = "12:30" + ++++ + +# openGauss+KeepAlived + +## 实验环境 + +操作系统： CentOS 7.6 + +数据库版本： openGauss 1.1.0Primary + +主机/IP： opengaussdb1/192.168.1.11 \(openGauss主备已部署完毕\) + +Standby 主机/IP： opengaussdb2/192.168.1.12 \(openGauss主备已部署完毕\) + +>![](public_sys-resources/icon-note.gif) **说明：** +>不建议在云环境\(如：华为云\)下搭建Keepalived进行测试，本人在云环境下测试发现，Keepalived的VIP无法在云环境下与其他主机通信，云环境下如何使用该VIP建议咨询云服务厂商。在踩坑之后，选择使用本地的VMWare workstation进行简单测试。 + +## 安装KeepAlived软件 + +``` +## 在所有节点执行安装 +yum install keepalived -y +``` + +## 配置keepalived + +>![](public_sys-resources/icon-note.gif) **说明：** +>采用nopreempt不抢占VIP，主备节点的state均设置为BACKUP。 + +- 主节点配置文件。 + +``` +# vi /etc/keepalived/keepalived.conf +-------------------------------------------- +! Configuration File for keepalived +## 全局定义 +global_defs { + router_id Keepalived_openGauss #运行 keepalived 服务器的一个标识 + script_user root #执行脚本的用户 +} + +## VRRP实例定义 +## 通常如果master服务Down掉后backup会变成master，但是当master服务又好了的时候 master此时会抢占VIP，这样就会发生两次数据库切换。 +## 建议使用nopreempt参数设置为非抢占模式，此时主库从故障中恢复后，不会从新的主库抢回VIP，但这需要将master和backup的state都设置成backup。 +vrrp_instance VI_1 { + state BACKUP #指定Keepalived的角色(BACKUP需大写) + interface eth0 #指定 HA 监测的网络接口 + virtual_router_id 59 #虚拟路由的数字标识，同一个 vrrp_instance 下，MASTER 和 BACKUP 一致 + nopreempt #非抢占模式，主库从故障中恢复后，不会从新的主库抢回VIP + priority 100 #优先级，备节点需要适当降低优先级 + advert_int 1 #MASTER 和 BACKUP 负载均衡器同步检查的时间间隔（秒） + authentication { #设置验证码和验证类型 + auth_type PASS + auth_pass 1111 + } + virtual_ipaddress { #设置虚拟 IP 地址，可以设置多个，每个一行 + 192.168.1.10 + } +} + +## 虚拟服务器定义 +virtual_server 192.168.1.10 26000 { #设置虚拟服务器的 IP 和端口，用空格隔开 + delay_loop 6 #设置运行情况检查时间，单位是秒 +# lb_algo rr #负载调度算法(轮询) +# lb_kind DR #负载均衡机制(NAT、TUN、DR) + persistence_timeout 50 #会话保持时间(秒) + protocol TCP #转发协议类型 + real_server 192.168.1.11 26000 { #配置服务节点 + weight 100 #配置服务节点的权重 + notify_down /gauss/failoverdb.sh #故障响应脚本 + TCP_CHECK { #使用TCP_CHECK方式进行健康检查 + connect_timeout 10 #10秒无响应即超时 + delay_before_retry 3 #重试间隔时间 + } + } +} +``` + +- 主节点故障切换脚本（仅适用openGauss进程崩溃故障处理，不适用Primary操作系统宕机故障处理）。 + +``` +vi /gauss/failoverdb.sh +-------------------------------------------- +#!/bin/bash +echo "Start to failover openGauss database." +pkill keepalived +ssh 192.168.1.12 "su - omm -c 'gs_ctl failover -D /gauss/data/db1'" +ssh 192.168.1.12 "su - omm -c 'gs_om -t refreshconf'" +echo 'Failover operation is completed.' +-------------------------------------------- +chmod 764 /gauss/failoverdb.sh +``` + +- 备节点配置文件。 + +``` +# vi /etc/keepalived/keepalived.conf +-------------------------------------------- +! Configuration File for keepalived +## 全局定义 +global_defs { + router_id Keepalived_openGauss #运行 keepalived 服务器的一个标识 + script_user root #执行脚本的用户 +} + +## VRRP实例定义 +## 通常如果master服务Down掉后backup会变成master，但是当master服务又好了的时候 master此时会抢占VIP，这样就会发生两次数据库切换。 +## 建议使用nopreempt参数设置为非抢占模式，此时主库从故障中恢复后，不会从新的主库抢回VIP，但这需要将master和backup的state都设置成backup。 +vrrp_instance VI_1 { + state BACKUP #指定Keepalived的角色(BACKUP需大写) + interface eth0 #指定 HA 监测的网络接口 + virtual_router_id 59 #虚拟路由的数字标识，同一个 vrrp_instance 下，MASTER 和 BACKUP 一致 + nopreempt #非抢占模式，主库从故障中恢复后，不会从新的主库抢回VIP + priority 60 #优先级，备节点需要适当降低优先级 + advert_int 1 #MASTER 和 BACKUP 负载均衡器同步检查的时间间隔（秒） + authentication { #设置验证码和验证类型 + auth_type PASS + auth_pass 1111 + } + virtual_ipaddress { #设置虚拟 IP 地址，可以设置多个，每个一行 + 192.168.1.10 + } +} + +## 虚拟服务器定义 +virtual_server 192.168.1.10 26000 { #设置虚拟服务器的 IP 和端口，用空格隔开 + delay_loop 6 #设置运行情况检查时间，单位是秒 +# lb_algo rr #负载调度算法(轮询) +# lb_kind DR #负载均衡机制(NAT、TUN、DR) + persistence_timeout 50 #会话保持时间(秒) + protocol TCP #转发协议类型 + real_server 192.168.1.12 26000 { #配置服务节点 + weight 60 #配置服务节点的权重 + notify_down /gauss/failoverdb.sh #虚拟服务故障响应脚本 + MISC_CHECK { ## 使用 MISC_CHECK 方式自定义脚本做健康检查 + misc_path "/gauss/check.sh" ## 检测脚本 + misc_timeout 10 ## 执行脚本的超时时间 + misc_dynamic ## 根据退出状态码动态调整服务器的权重 + } + } +} +-------------------------------------------- +## 备节点选择MISC_CHECK方式的原因： +## 测试发现，当主节点直接断电宕机后，Keepalived的VIP会漂移至备节点，此时如果使用TCP_CHECK方式做健康检查，会因为备机可读的原因使得VIP:26000连接正常，造成keepalived健康检查的误判。 +## 最终导致主节点断电宕机后，备节点虽获取了VIP，但并没有执行openGauss的failover操作，备节点依旧只读，无法对外提供业务。 +## 为了纠正这一点，建议使用MISC_CHECK方式自定义脚本，登录主节点做数据库健康检查(简单示例脚本：/gauss/check.sh) +``` + +- 备节点健康检查脚本\[ ssh登录主节点进行数据库连接检查 \]。 + +``` +vi /gauss/check.sh +------------------------------------------- +ssh 192.168.1.11 "su - omm -c \"gsql -d postgres -p 26000 -t -A -c 'select 1;'\"" +------------------------------------------- +``` + +- 备节点故障切换脚本。 + +``` +vi /gauss/failoverdb.sh +-------------------------------------------- +#!/bin/bash +echo "Start to failover openGauss database." +pkill keepalived +su - omm -c "gs_ctl failover -D /gauss/data/db1" +su - omm -c "gs_om -t refreshconf" +echo 'Failover operation is completed.' +-------------------------------------------- +chmod 764 /gauss/failoverdb.sh +``` + +## openGauss配置 + +- 修改openGauss侦听地址。 + +``` +$ gs_guc set -I all -N all -c "listen_addresses = '0.0.0.0'" +$ gs_guc set -I all -N all -c "local_bind_address = '0.0.0.0'" +``` + +- 修改所有节点replconninfo参数（避免端口冲突）。 + +``` +$ vi /gauss/data/db1/postgresql.conf +-------------------------------------------- +修改：localport --> 26011 +修改：remoteport --> 26011 +-------------------------------------------- +``` + +- 重启openGauss数据库,并检查服务器状态。 + +``` +## 重启openGauss +[omm@prod db1]$ gs_om -t stop && gs_om -t start + +## 检查openGauss状态 +[root@opengaussdb1 ~]# su - omm -c "gs_om -t status --detail" +[ Cluster State ] +cluster_state : Normal +redistributing : No +current_az : AZ_ALL +[ Datanode State ] +node node_ip instance state | +----------------------------------------------------------------------- +1 opengaussdb1 192.168.1.11 6001 /gauss/data/db1 P Primary Normal | +2 opengaussdb2 192.168.1.12 6002 /gauss/data/db1 S Standby Normal + +## 检查KeepAlived进程状态 +[omm@opengaussdb1 ~]$ ps -ef|grep keep|grep -v grep +root 15664 1 0 16:15 ? 00:00:00 /usr/sbin/keepalived -D +root 15665 15664 0 16:15 ? 00:00:00 /usr/sbin/keepalived -D +root 15666 15664 0 16:15 ? 00:00:00 /usr/sbin/keepalived -D + +## 检查VIP状态 +[root@opengaussdb1 ~]# ip a +1: lo: mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000 + link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 + inet 127.0.0.1/8 scope host lo + valid_lft forever preferred_lft forever + inet6 ::1/128 scope host + valid_lft forever preferred_lft forever +2: ens33: mtu 1500 qdisc pfifo_fast state UP group default qlen 1000 + link/ether 00:0c:29:da:60:c0 brd ff:ff:ff:ff:ff:ff + inet 192.168.1.11/24 brd 192.168.1.255 scope global noprefixroute ens33 + valid_lft forever preferred_lft forever + inet 192.168.1.10/32 scope global ens33 ## VIP：192.168.1.10 + valid_lft forever preferred_lft forever + inet6 2408:8270:237:ded0:c89c:adab:e7b:8bd6/64 scope global noprefixroute dynamic + valid_lft 258806sec preferred_lft 172406sec + inet6 fe80::c4f2:8ad1:200d:ce9b/64 scope link noprefixroute + valid_lft forever preferred_lft forever +``` + +## 故障模拟测试 + +- 主节点\[192.168.1.11\]操作。 + +``` +## kill数据库进程 +[root@opengaussdb1 ~]# ps -ef|grep gauss +omm 18115 1 4 16:30 ? 00:00:35 /gauss/app/bin/gaussdb -D /gauss/data/db1 -M primary +root 19254 9299 0 16:42 pts/0 00:00:00 grep --color=auto gauss +[root@opengaussdb1 ~]# kill -9 18115 + +## 检查message日志[检测到故障，执行notify_down脚本，并关闭keepalived服务] +# tail -fn 200 /var/log/messages +Feb 19 16:42:57 opengaussdb1 Keepalived_healthcheckers[18816]: TCP connection to [192.168.1.11]:26000 failed. +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: TCP connection to [192.168.1.11]:26000 failed. +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: Check on service [192.168.1.11]:26000 failed after 1 retry. +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: Removing service [192.168.1.11]:26000 from VS [192.168.1.10]:26000 +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: IPVS (cmd 1160, errno 2): No such destination +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: Executing [/gauss/failoverdb.sh] for service [192.168.1.11]:26000 in VS [192.168.1.10]:26000 +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: Lost quorum 1-0=1 > 0 for VS [192.168.1.10]:26000 +Feb 19 16:43:00 opengaussdb1 Keepalived[18815]: Stopping +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: pid 19258 exited due to signal 15 +Feb 19 16:43:00 opengaussdb1 Keepalived_vrrp[18817]: VRRP_Instance(VI_1) sent 0 priority +Feb 19 16:43:00 opengaussdb1 Keepalived_vrrp[18817]: VRRP_Instance(VI_1) removing protocol VIPs. +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: IPVS (cmd 1156, errno 2): No such file or directory +Feb 19 16:43:00 opengaussdb1 Keepalived_healthcheckers[18816]: Stopped +Feb 19 16:43:01 opengaussdb1 Keepalived_vrrp[18817]: Stopped +Feb 19 16:43:01 opengaussdb1 Keepalived[18815]: Stopped Keepalived v1.3.5 (03/19,2017), git commit v1.3.5-6-g6fa32f2 +``` + +- 备节点\[192.168.1.12\]检查。 + +``` +## 检查VIP是否已漂移 +[root@opengaussdb2 ~]# ip a|grep 192.168 + inet 192.168.1.12/24 brd 192.168.1.255 scope global noprefixroute ens33 + inet 192.168.1.10/32 scope global ens33 + +## 检查数据库状态[已failover成为Primary] +[omm@opengaussdb2 ~]$ gs_om -t status --detail +[ Cluster State ] +cluster_state : Degraded +redistributing : No +current_az : AZ_ALL +[ Datanode State ] +node node_ip instance state | +--------------------------------------------------------------------------------- +1 opengaussdb1 192.168.1.11 6001 /gauss/data/db1 P Down Manually stopped | +2 opengaussdb2 192.168.1.12 6002 /gauss/data/db1 S Primary Normal +``` + -- Gitee