【经验分享】openGauss容灾集群搭建

gs_sdr命令代码解读

背景

openGauss推出了容灾架构，相比之前的一个集群主从架构，而容灾架构是两个集群间的数据同步。为了更深入了解其原理，本文试图通过阅读gs_sdr命令相关的代码来学习下相关的各种操作。

1.容灾搭建过程可以参考：https://www.modb.pro/db/628767

2.vscode调试配置可以参考：https://www.modb.pro/db/658344

3.个人学习记录，理解不一定完全正确。如有错误，可指出一起探讨_

环境准备

安装集群

安装两套集群，每套集群含2个节点，相关信息如下：

集群1信息

omm@pghost2 ~$ cm_ctl query -Cvid
[ CMServer State ]

node node_ip instance state
---------------------------------------------------------------------
1 pghost2 192.168.56.20 1 /app/ogdata/data/cm/cm_server Primary
2 pghost3 192.168.56.30 2 /app/ogdata/data/cm/cm_server Standby

[ Cluster State ]

cluster_state : Normal
redistributing : No
balanced : Yes
current_az : AZ_ALL

[ Datanode State ]

node node_ip instance state | node node_ip instance state
------------------------------------------------------------------------------------------------------------------------------------------------
1 pghost2 192.168.56.20 6001 /app/ogdata/data/dn1 P Primary Normal | 2 pghost3 192.168.56.30 6002 /app/ogdata/data/dn1 S Standby Normal

集群2信息

omm@pghost5 ~$ cm_ctl query -Cvid
[ CMServer State ]

node node_ip instance state
---------------------------------------------------------------------
1 pghost5 192.168.56.50 1 /app/ogdata/data/cm/cm_server Primary
2 pghost6 192.168.56.60 2 /app/ogdata/data/cm/cm_server Standby

[ Cluster State ]

cluster_state : Normal
redistributing : No
balanced : Yes
current_az : AZ_ALL

[ Datanode State ]

node node_ip instance state | node node_ip instance state
------------------------------------------------------------------------------------------------------------------------------------------------
1 pghost5 192.168.56.50 6001 /app/ogdata/data/dn1 P Primary Normal | 2 pghost6 192.168.56.60 6002 /app/ogdata/data/dn1 S Standby Normal

创建容灾用户

在集群1上创建容灾用户:

gsql -d postgres -p 26000 -c "create user dr_user with replication password 'oracle_4U';"

修改XML配置

修改集群1

修改后的xml配置如下：

> > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > **** > **** > > > > > > > > > > > > > > > > > > > >

修改集群2

修改后的xml配置如下：

> > >   >     >     >     >     >     >     >     >     >     >   >   >     >       >       >       >       >       >       >       >       >       >       >       >       >       >       >       >       >       **** >       **** >       **** >     >     >       >       >       >       >       >       >     >   > 配置容灾

将集群1启动为主集群

使用的命令为：

# gs_sdr -t start -m primary -X XMLFILE [-U DR_USERNAME [-W DR_PASSWORD]] [--time-out=SECS]
gs_sdr -t start -m primary -X /home/omm/single.xml -U dr_user -W oracle_4U --time-out=86400

vscode调试配置

{
"version": "0.2.0",
"configurations": [
{
"name": "Python: 当前文件",
"type": "python",
"request": "launch",
"program": "${file}",
"console": "integratedTerminal",
"justMyCode": true,
"args": ["-t","start","-m","primary","-X","/home/omm/single.xml","-U","dr_user","-W","oracle_4U","--time-out=86400"]
}
]
}

在gs_sdr脚本main函数中打上断点

代码阅读

判断是否使用root权限操作

if os.getuid() == 0:
GaussLog.exitWithError(ErrorCode.GAUSS_501["GAUSS_50105"])
# 是root权限就直接报错退出

初始化StreamingDisasterRecoveryBase类

base = StreamingDisasterRecoveryBase() # 从集群xml配置文件中加载相 关的信息

base中保存的信息可以参考下图：

判断做何种操作

handler = HANDLER_MAPPING[base.params.task](base.params, base.user, base.logger, base.trace_id, base.log_file)
# 这里的 HANDLER_MAPPING 主要包括4种操作。具体如下：
HANDLER_MAPPING = {
"start": StreamingStartHandler, # 这块应该是对应上图中的 moduleName 中的值
"stop": StreamingStopHandler,
"switchover": StreamingSwitchoverHandler,
"failover": StreamingFailoverHandler,
"query": StreamingQueryHandler
}
# 此处的 base.params.task 值为 start ，映射到类 StreamingStartHandler ，该类在文件 streaming_diaster_recovery_start.py 中

创建锁定文件

由于容灾搭建过程涉及到数据同步耗时较长，这里应是为避免多次重复操作。

handler.handle_lock_file(handler.trace_id, 'create') # 该方法在streaming_base.py中定义
# 会生成一个文件：'/app/opengauss/tmp/streaming_lock_cd7eef1a2c1f11ee92b208002716c96f'

判断是否有其他gs_sdr操作

if base.params.task in StreamingConstants.TASK_EXIST_CHECK:
handler.check_streaming_process_is_running() # 有的话，就终止本次操作。
# 'source /home/omm/.bashrc && pssh -t 10 -H pghost2 -H pghost3 "ls /app/opengauss/tmp/streaming_lock_*"' 主要使用该命令

执行操作

进度记录相关操作

handler.run()
self.logger.log("Start create streaming disaster relationship.")
# 创建进度记录文件夹：/app/opengauss/tmp/streaming_cabin(所有节点均创建)
# 进度记录文件：'.streaming_switchover_primary.step'
## 所有的进度记录文件名字如下：
STREAMING_STEP_FILES = {
"start_primary": ".streaming_start_primary.step",
"start_standby": ".streaming_start_standby.step",
"stop": ".streaming_stop.step",
"switchover_primary": ".streaming_switchover_primary.step",
"switchover_standby": ".streaming_switchover_standby.step",
"failover": ".streaming_failover.step",
"query": ".streaming_query.step",
}

检查集群状态

# 检查集群状态
'source /home/omm/.bashrc ; gs_om -t status --all > /app/opengauss/tmp/streaming_cabin/cluster_state_tmp'

判断执行节点是否为主节点

操作需要在主节点上执行。

生成 key_name.key.cipher & key_name.key.rand 文件

export LD_LIBRARY_PATH=/app/opengauss/tool/script/gspylib/clib && source /home/omm/.bashrc && gs_guc generate -S default -o hadr -D '/app/opengauss/app/2.0.1_46134f73/bin' && /bin/chmod 600 /app/opengauss/app/2.0.1_46134f73/bin/hadr.key.cipher && /bin/chmod 600 /app/opengauss/app/2.0.1_46134f73/bin/hadr.key.rand
# 随后会将生成的文件分发到集群中其他节点上。

保存hadr信息到数据库

ALTER GLOBAL CONFIGURATION with(hadr_user_info ='O1hnmUERtm2hfiXGjKjgaCfKq89IgdSzUqCoMGw/yzdaYki1LYTfhHlILmz10IvDTX9fqGNZrcmdX5NmkK+6bw==');

检查是否已经有首备节点

判断是否已经是容灾环境。

检查是否有cm

容灾环境必须要有cm组件。

检查是否在升级中

# 判断/app/opengauss/tmp/binary_upgrade是否存在

写进度文件

$ more /app/opengauss/tmp/streaming_cabin/.streaming_start_primary.step
2_check_cluster_step

common_step_for_streaming_start

# 生成容灾关系json文件 并分发到集群中的其它节点上。
more /app/opengauss/tmp/streaming_cabin/cluster_conf_record
{"remoteClusterConf": {"port": 26500, "shards": [[{"ip": "192.168.56.50", "dataIp": "192.168.56.50"}, {"ip": "1
92.168.56.60", "dataIp": "192.168.56.60"}]]}, "localClusterConf": {"port": 26000, "shards": [[{"ip": "192.168.5
6.20", "dataIp": "192.168.56.20"}, {"ip": "192.168.56.30", "dataIp": "192.168.56.30"}]]}}

修改pg_hba配置

# 拷贝/home/omm/single.xml为/app/opengauss/tmp/streaming_cabin/streaming_config.xml
source /home/omm/.bashrc; python3 '/app/opengauss/tool/script/local/ConfigHba.py' -U omm -X '/app/opengauss/tmp/streaming_cabin/streaming_config.xml' --try-reload
# 会在pg_hba.conf文件中加入：
host all omm 192.168.56.50/32 trust
host all omm 192.168.56.60/32 trust
host replication all 192.168.0.0/16 sha256

复制参数replconninfo相关设置

'source /home/omm/.bashrc; pssh -H pghost3 \'source /home/omm/.bashrc; gs_guc check -Z datanode -D /app/ogdata/data/dn1 -c "replconninfo1"\''

'source /home/omm/.bashrc; pssh -H pghost3 \'source /home/omm/.bashrc; gs_guc check -Z datanode -D /app/ogdata/data/dn1 -c "replconninfo2"\''

'source /home/omm/.bashrc; pssh -H pghost3 "source /home/omm/.bashrc ; gs_guc reload -Z datanode -D /app/ogdata/data/dn1 -c \\"replconninfo1 = \'localhost=192.168.56.30 localport=26001 localheartbeatport=26005 localservice=26004 remotehost=192.168.56.20 remoteport=26001 remoteheartbeatport=26005 remoteservice=26004 iscascade=true iscrossregion=false\'\\""'

等待首备连接

Waiting for the main standby connection.

这里需要在备集群执行下面的命令：

gs_sdr -t start -m disaster_standby -U dr_user -W oracle_4U -X /home/omm/single.xml --time-out=86400 # 此处为方便，直接在终端上执行该命令，没有进行调试。

将集群2启动为备集群

gs_sdr -t start -m disaster_standby -U dr_user -W oracle_4U -X /home/omm/single.xml --time-out=86400

vscode调试配置

{
"version": "0.2.0",
"configurations": [
{
"name": "gs_sdr",
"type": "python",
"request": "launch",
"program": "${file}",
"console": "integratedTerminal",
"justMyCode": true,
"args": ["-t","start","-m","disaster_standby","-X","/home/omm/single.xml","-U","dr_user","-W","oracle_4U","--time-out=86400"]
}
]
}

执行的类: streaming_diaster_recovery_start

代码阅读

Start build key files from remote cluster

备集群会进行build，速度比较慢(与网络环境和数据库大小关系较大)。

source /home/omm/.bashrc; /app/opengauss/app/2.0.1/bin/gs_ctl build -D /app/ogdata/data/dn1 -M standby -b copy_secure_files -U dr_user -P *** -C "localhost=192.168.56.50 localport=26001 remotehost=192.168.56.20 remoteport=26501"

source /home/omm/.bashrc; /app/opengauss/app/2.0.1/bin/gs_ctl build -D /app/ogdata/data/dn1 -M standby -b copy_secure_files -U dr_user -P *** -C "localhost=192.168.56.50 localport=26001 remotehost=192.168.56.30 remoteport=26501"

echo *** /home/omm/.bashrc; /app/opengauss/app/2.0.1/bin/gs_ctl build -D /app/ogdata/data/dn1 -M standby -b copy_secure_files -U dr_user -P *** -C 'localhost=192.168.56.60 localport=26001 remotehost=192.168.56.20 remoteport=26501'" | pssh -s -H pghost6'

copy file from data dir to streaming dir

# 第1个节点
echo "if [ -d \'/app/ogdata/data/dn1/gs_secure_files\' ];then source /home/omm/.bashrc && pscp --trace-id 9f2c898e2c5a11ee850c080027fd3332 -H pghost5 \'/app/ogdata/data/dn1/gs_secure_files\' \'/app/opengauss/tmp/streaming_cabin\' && rm -rf \'/app/ogdata/data/dn1/gs_secure_files\';fi" | pssh -s -H pghost5
# 第2个节点
echo "if [ -d \'/app/ogdata/data/dn1/gs_secure_files\' ];then source /home/omm/.bashrc && pscp --trace-id 9f2c898e2c5a11ee850c080027fd3332 -H pghost5 \'/app/ogdata/data/dn1/gs_secure_files\' \'/app/opengauss/tmp/streaming_cabin\' && rm -rf \'/app/ogdata/data/dn1/gs_secure_files\';fi" | pssh -s -H pghost6

check cluster user consistency

主要检查版本和版本提交号是否一致。

检查安装用户是否一致

设置集群运行模式stream_cluster_run_mode

source /home/omm/.bashrc && gs_guc set -Z datanode -N all -I all -c "stream_cluster_run_mode = \'cluster_standby\'"

source /home/omm/.bashrc && gs_guc set -Z coordinator -N all -I all -c "stream_cluster_run_mode = \'cluster_standby\'"

停止备集群

'/app/opengauss/app/2.0.1_46134f73/bin/cluster_static_config'

再次build集群

source /home/omm/.bashrc; /app/opengauss/app/2.0.1_46134f73/bin/gs_ctl start -D /app/ogdata/data/dn1 -M hadr_main_standby

echo *** /home/omm/.bashrc; /app/opengauss/app/2.0.1_46134f73/bin/gs_ctl build -D /app/ogdata/data/dn1 -M cascade_standby -b standby_full -r 7200 -t 1209600" | pssh -s -t 1209610 -H pghost6

启动集群

source /home/omm/.bashrc ; cm_ctl start -t 604800 # 此时的集群已经是首备和级联备状态了。

查询容灾状态

gs_sdr -t query

主集群

$ gs_sdr -t query
--------------------------------------------------------------------------------
Streaming disaster recovery query 9f658f3a2d0511eebbb208002716c96f
--------------------------------------------------------------------------------
Start streaming disaster query.
Start check archive.
Start check recovery.
Start check RPO & RTO.
Successfully executed streaming disaster recovery query, result:
{'hadr_cluster_stat': 'archive', 'hadr_failover_stat': '', 'hadr_switchover_stat': '', 'RPO': '0', 'RTO': '0'}

备集群

$ gs_sdr -t query
--------------------------------------------------------------------------------
Streaming disaster recovery query ad8afd5c2d0511ee88cf080027fd3332
--------------------------------------------------------------------------------
Start streaming disaster query.
Start check archive.
Start check recovery.
Start check RPO & RTO.
Successfully executed streaming disaster recovery query, result:
{'hadr_cluster_stat': 'restore', 'hadr_failover_stat': '', 'hadr_switchover_stat': '', 'RPO': '', 'RTO': ''}