前文
- 假定您对Azure和SQL Server HA具有基础知识
- 假定您对Azure Cli具有基础知识
- 目标是在Azure Linux VM上创建一个具有三个副本的可用性组,并实现侦听器和Fencing配置
环境
- SQL Server 2019 Developer on Linux
- Azure VM Fencing agent
- Azure Cli实现部分配置
- CentOS 7.7 Azure VM,分别SQL19N1,SQL19N2,SQL19N3,位于同一VNet
步骤
为VM创建资源组和可用性集
# 中国东部2创建资源组
az group create --name SQL-DEMO-RG --location chinaeast2
# 创建用于VM的Availability Set,配置2个容错域,2个更新域
az vm availability-set create \
--resource-group SQL-DEMO-RG \
--name AGLinux-AvailabilitySet \
--platform-fault-domain-count 2 \
--platform-update-domain-count 2
使用Template部署3台VM
第一次创建VM时,会生成template,然后下载保存下,修改其中的参数值后,就可以方便地创建配置类似的VM。VM的配置主要有:
- 使用前面的可用性集
- 使用同一个子网
- IP使用Standard
- SSH public key配置
模板和参数文件太长,就不展示了。可以在Azure Portal上自行获取。
# 如下是SQL19N2的配置,修改参数文件后,直接可以用于创建SQL19N3
templateFile="./templateFile"
paramFile="./vmParams-sql19n2.json"
az deployment group validate --name sql19n2vm \
-g SQL-DEMO-RG --template-file $templateFile --parameters $paramFile
配置VM使用固定内网IP和公网DNS Label
三台VM都需要修改配置,如下只是一台的配置示例
# 找出nic和IP的信息
az network nic list -g SQL-DEMO-RG --query "[].{nicName:name,configuration:ipConfigurations[].{ipName:name,ip:privateIpAddress,method:privateIpAllocationMethod}}" -o yaml
# 修改privateIpAllocationMethod为Static
az network nic ip-config update -g SQL-DEMO-RG --nic-name sql19n1152 --name ipconfig1 --set privateIpAllocationMethod=Static
# 找出pbulic ip名称
az network public-ip list -g SQL-DEMO-RG --query "[].name" -o tsv
# 配置Public IP的DNS name,只能使用数字和小写字母
az network public-ip update -g SQL-DEMO-RG -n SQL19N1ip851 --dns-name sql19n1
安装HA相关软件包
最好先更新一下系统的软件包,再安装HA相关软件。
yum update -y
yum install -y pacemaker pcs fence-agents-all resource-agents fence-agents-azure-arm
reboot
为群集和SQL Server开放防火墙端口
# Pacemaker和Corosync的端口
# TCP: Ports 2224,3121,21064,5405
# UDP: Port 5405
firewall-cmd --add-port=2224/tcp --permanent
firewall-cmd --add-port=2224/tcp --permanent
firewall-cmd --add-port=21064/tcp --permanent
firewall-cmd --add-port=5405/tcp --permanent
firewall-cmd --add-port=5405/udp --permanent
# SQL Server端口和AG镜像端口
# TCP: 1433,5022
firewall-cmd --add-port=1433/tcp --permanent
firewall-cmd --add-port=5022/tcp --permanent
firewall-cmd --reload
添加hosts记录
vi /etc/hosts
172.17.2.8 SQL19N1
172.17.2.9 SQL19N2
172.17.2.10 SQL19N3
创建Pacemaker群集
# 设置Pacemaker的默认用户密码,三台VM上
passwd hacluster
# 设置pacemaker和pcsd自启动在三台VM上
systemctl enable pcsd
systemctl start pcsd
systemctl enable pacemaker
# 创建群集,在master节点
sudo pcs cluster auth SQL19N1 SQL19N2 SQL19N3 -u hacluster
sudo pcs cluster setup --name agcluster SQL19N1 SQL19N2 SQL19N3 --token 30000 --force
sudo pcs cluster start --all
sudo pcs cluster enable --all
# 查看群集状态
pcs status
# 在三个节点上修改quorum的expected-votes为3,其实三节点群集默认为3
# 设置表示,群集存活需要3票,这个修改只影响当前running的群集,不会变成群集的永久性配置保存下来
pcs quorum expected-votes 3
在Azure上为Fencing Agent配置Service Principal
# 1. 创建 AAD App,成功后记录下相应的appID
az ad app create --display-name sqldemorg-app --identifier-uris http://localhost \
--password "1qaz@WSX3edc" --end-date '2030-04-27' --credential-description "sql19 ag secret"
# 2. 创建AAD App的Service Principal
az ad sp create --id <appID>
# 3. 将Service Principal分配到VM对应的管理role,对每个VM都要执行
# 我这里分配的是Owner role,这不是安全的做法。应该使用自定义一个role,只给最小权限
# 自定义role需要Azure订阅是PP1或者PP2级别
az role assignment create --assignee <appID> --role owner \
--scope /subscriptions/<subscription-ID>/resourceGroups/<resourceGroup-Name>/providers/Microsoft.Compute/virtualMachines/SQL19N1
创建Azure的STONITH 设备
我使用的是Azure China,所以需要指定cloud=china,如果使用global Azure不需要指定此参数。
执行 fence_azure_arm -h,查看此资源代理的更多帮助信息
pcs property set stonith-timeout=900
pcs stonith create rsc_st_azure fence_azure_arm login="<ApplicationID>" passwd="<servicePrincipalPassword>" resourceGroup="<resourceGroupName>" tenantId="<tenantID>" subscriptionId="<subscriptionId>" power_timeout=240 pcmk_reboot_timeout=900 cloud=china
安装SQL 2019及工具
# 安装 SQL 2019和HA 资源代理
sudo curl -o /etc/yum.repos.d/mssql-server.repo https://packages.microsoft.com/config/rhel/7/mssql-server-2019.repo
sudo yum install -y mssql-server
sudo /opt/mssql/bin/mssql-conf setup
sudo yum install mssql-server-ha
# 安装 mssql-tools
sudo curl -o /etc/yum.repos.d/msprod.repo https://packages.microsoft.com/config/rhel/7/prod.repo
sudo yum install -y mssql-tools unixODBC-devel
# 将mssql-tools目录加入到aPATH,方便使用
echo 'export PATH="$PATH:/opt/mssql-tools/bin"' >> ~/.bash_profile
echo 'export PATH="$PATH:/opt/mssql-tools/bin"' >> ~/.bashrc
source ~/.bashrc
# 安装 mssql-cli
sudo rpm --import https://packages.microsoft.com/keys/microsoft.asc
sudo curl -o /etc/yum.repos.d/mssql-cli.repo https://packages.microsoft.com/config/rhel/7/prod.repo
sudo yum install mssql-cli
# 查看SQL 状态
systemctl status mssql-server
如果您熟悉 SQL Server相关的PowerShell,建议将PowerShell也安装上,并安装SQLServer module。对SQL Server的配置,使用PowerShell会方便很多
yum install powershell -y
pwsh
Install-Module SQLServer
# 查看SQL相关的命令
Get-Command -Module SQLServer
配置AG
创建PowerShell 函数方便后续执行T-SQL
# 打开PowerShell的 profile文件,如果不存在则需要创建
vi /root/.config/powershell/Microsoft.PowerShell_profile.ps1
# 将如下函数加入 到 profile文件中,每次打开pwsh时就可以直接调用
# 函数有两个参数,$sql表示需要执行的T-SQL,最好使用here-string以避免字符转义问题
# $servers表示目标实例,数组类型。默认值为当前环境中的三个实例
function run-sql ($sql,$servers=("SQL19N1","SQL19N2","SQL19N3"))
{
$secpasswd = "1qaz@WSX"|ConvertTo-SecureString -AsPlainText -Force
$cred=New-Object System.Management.Automation.PSCredential -ArgumentList 'sa', $secpasswd
$sql
"---------"
foreach($svr in $servers) {"Running T-SQL on $svr..."; Invoke-Sqlcmd -ServerInstance $svr -Credential $cred -Query $sql}
}
启用 hadr功能,每个实例
sudo /opt/mssql/bin/mssql-conf set hadr.hadrenabled 1
sudo systemctl restart mssql-server
启动AG extened event session
# T-SQL,每个实例
ALTER EVENT SESSION AlwaysOn_health ON SERVER WITH (STARTUP_STATE=ON);
GO
在主副本实例上创建证书,这个证书用于验证Mirroring endpoint通信。将证书和私钥复制到其它节点上的相同的目录位置。授予mssql用户访问权限
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '1qaz@WSX';
GO
CREATE CERTIFICATE dbm_certificate WITH SUBJECT = 'dbm';
GO
BACKUP CERTIFICATE dbm_certificate
TO FILE = '/var/opt/mssql/data/dbm_certificate.cer'
WITH PRIVATE KEY (
FILE = '/var/opt/mssql/data/dbm_certificate.pvk',
ENCRYPTION BY PASSWORD = '1qaz@WSX'
);
# 复制证书和私钥到辅助副本主机SQL19N2和SQL19N3
cd /var/opt/mssql/data
scp dbm_certificate.* root@SQL19N2:/var/opt/mssql/data/
scp dbm_certificate.* root@SQL19N3:/var/opt/mssql/data/
# 辅助副本节点上修改权限
cd /var/opt/mssql/data
chown mssql:mssql dbm_certificate.*
在辅助副本实例中创建master key并导入证书
CREATE MASTER KEY ENCRYPTION BY PASSWORD = '1qaz@WSX';
GO
CREATE CERTIFICATE dbm_certificate
FROM FILE = '/var/opt/mssql/data/dbm_certificate.cer'
WITH PRIVATE KEY (
FILE = '/var/opt/mssql/data/dbm_certificate.pvk',
DECRYPTION BY PASSWORD = '1qaz@WSX'
);
创建AG的镜像端口,注意防火墙和NSG配置端口例外
CREATE ENDPOINT [Hadr_endpoint]
AS TCP (LISTENER_PORT = 5022)
FOR DATABASE_MIRRORING (
ROLE = ALL,
AUTHENTICATION = CERTIFICATE dbm_certificate,
ENCRYPTION = REQUIRED ALGORITHM AES
);
GO
ALTER ENDPOINT [Hadr_endpoint] STATE = STARTED;
创建三个副本,同步模式的AG,主副本实例上执行
CREATE AVAILABILITY GROUP [ag1]
WITH (DB_FAILOVER = ON, CLUSTER_TYPE = EXTERNAL)
FOR REPLICA ON
N'SQL19N1'
WITH (
ENDPOINT_URL = N'tcp://SQL19N1:5022',
AVAILABILITY_MODE = SYNCHRONOUS_COMMIT,
FAILOVER_MODE = EXTERNAL,
SEEDING_MODE = AUTOMATIC,
SECONDARY_ROLE(ALLOW_CONNECTIONS = ALL)
),
N'SQL19N2'
WITH (
ENDPOINT_URL = N'tcp://SQL19N2:5022',
AVAILABILITY_MODE = SYNCHRONOUS_COMMIT,
FAILOVER_MODE = EXTERNAL,
SEEDING_MODE = AUTOMATIC,
SECONDARY_ROLE(ALLOW_CONNECTIONS = ALL)
),
N'SQL19N3'
WITH(
ENDPOINT_URL = N'tcp://SQL19N3:5022',
AVAILABILITY_MODE = SYNCHRONOUS_COMMIT,
FAILOVER_MODE = EXTERNAL,
SEEDING_MODE = AUTOMATIC,
SECONDARY_ROLE(ALLOW_CONNECTIONS = ALL)
);
GO
ALTER AVAILABILITY GROUP [ag1] GRANT CREATE ANY DATABASE;
GO
为Pacemaker创建sql登录并授权,每个实例
USE [master]
GO
CREATE LOGIN [pacemakerLogin] with PASSWORD= N'1qaz@WSX'
go
ALTER SERVER ROLE [sysadmin] ADD MEMBER [pacemakerLogin];
GO
将pacemaker的login信息保存到本地文件
echo "pacemakerLogin" >> /var/opt/mssql/secrets/passwd
echo "1qaz@WSX" >> /var/opt/mssql/secrets/passwd
# 只允许root读取
chown root:root /var/opt/mssql/secrets/passwd
chmod 400 /var/opt/mssql/secrets/passwd
将辅助副本加入到AG, 辅助副本执行
ALTER AVAILABILITY GROUP [ag1] JOIN WITH (CLUSTER_TYPE = EXTERNAL);
GO
# auto_seeding功能需要的权限
ALTER AVAILABILITY GROUP [ag1] GRANT CREATE ANY DATABASE;
GO
如果您不希望pacemakerLogin具有sysadmin的权限,可以将之从sysadmin中移除,并授予如下权限。每个实例
ALTER SERVER ROLE [sysadmin] DROP MEMBER [pacemakerLogin]
GO
GRANT ALTER, CONTROL, VIEW DEFINITION ON AVAILABILITY GROUP::ag1 TO pacemakerLogin;
GO
GRANT VIEW SERVER STATE TO pacemakerLogin;
GO
添加数据库到AG,主副本执行
CREATE DATABASE [db1];
GO
ALTER DATABASE [db1] SET RECOVERY FULL;
GO
BACKUP DATABASE [db1]
TO DISK = N'nul';
GO
ALTER AVAILABILITY GROUP [ag1] ADD DATABASE [db1];
GO
可用性数据库状态
SELECT * FROM sys.databases WHERE name = 'db1';
GO
SELECT DB_NAME(database_id) AS 'database', synchronization_state_desc FROM sys.dm_hadr_database_replica_states;
在Pacemaker群集中配置AG
创建AG资源,ag_name要指定为之前创建AG名称
pcs resource create agcluster ocf:mssql:ag ag_name=ag1 meta failure-timeout=30s master notify=true
创建虚拟IP资源
# 禁用fencing
pcs property set stonith-enabled=false
# 创建VIP
pcs resource create virtualip ocf:heartbeat:IPaddr2 ip=172.17.2.7
创建 colacation constraint,vip和master必需在同一个节点上启动
pcs constraint colocation add virtualip agcluster-master INFINITY with-rsc-role=Master
创建 ordering constraint,vip要先于master副本资源启动
pcs constraint order promote agcluster-master then start virtualip
# 查看当前的约束
pcs constraint show --full
重新启用STONITH并查看群集状态
pcs property set stonith-enabled=true
pcs status
# 我的环境中的状态信息
---------------------------------------
Cluster name: agcluster
Stack: corosync
Current DC: SQL19N3 (version 1.1.20-5.el7_7.2-3c4c782f70) - partition with quorum
Last updated: Wed Apr 29 04:24:50 2020
Last change: Wed Apr 29 04:24:45 2020 by root via cibadmin on SQL19N1
3 nodes configured
5 resources configured
Online: [ SQL19N1 SQL19N2 SQL19N3 ]
Full list of resources:
rsc_st_azure (stonith:fence_azure_arm): Started SQL19N1
Master/Slave Set: agcluster-master [agcluster]
Masters: [ SQL19N1 ]
Slaves: [ SQL19N2 SQL19N3 ]
virtualip (ocf::heartbeat:IPaddr2): Started SQL19N1
Daemon Status:
corosync: active/enabled
pacemaker: active/enabled
pcsd: active/enabled
测试Failover和Fencing
# 手动failover
pcs resource move agcluster-master SQL19N2 --master
pcs status
# 手动 failover会生成一个constraint,避免AG资源再回到原来的节点
# 如果希望AG后续还能 failover回来,需要手动删除之
pcs constraint show --full
pcs constraint remove cli-prefer-agcluster-master
# 尝试Fencing群集节点,每个节点都试一下
# 如下命令的fencing只是重启node,如果要关闭node,使用--off参数
pcs stonith fence SQL19N3 --debug