1.准备
本次针对GP6.x版本之前
主备节点：
master:node01
standbymaster:node02
查看集群备用节点状态：
gpstate -f
20190104:14:34:34:261113 gpstate:node01:gpadmin-[INFO]:-Starting gpstate with args: -f
20190104:14:34:34:261113 gpstate:node01:gpadmin-[INFO]:-local Greenplum Version: 'postgres (Greenplum Database) 5.14.0 build commit:13927a46a9b39527fc1af2cbb1571d4fd394ff88'
20190104:14:34:34:261113 gpstate:node01:gpadmin-[INFO]:-master Greenplum Version: 'PostgreSQL 8.3.23 (Greenplum Database 5.14.0 build commit:13927a46a9b39527fc1af2cbb1571d4fd394ff88) on x86_64-pc-linux-gnu, compiled by GCC gcc (GCC) 6.2.0, 64-bit compiled on Nov 28 2018 08:08:06'
20190104:14:34:34:261113 gpstate:node01:gpadmin-[INFO]:-Obtaining Segment details from master...
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:-Standby master details
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:-----------------------
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:- Standby address = node02
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:- Standby data directory = /data/master/gpseg-1
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:- Standby port = 5432
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:- Standby PID = 139960
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:- Standby status = Standby host passive
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--------------------------------------------------------------
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--pg_stat_replication
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--------------------------------------------------------------
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--WAL Sender State: streaming
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--Sync state: sync
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--Sent Location: 0/199DF208
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--Flush Location: 0/199DF208
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--Replay Location: 0/199DF208
20190104:14:34:36:261113 gpstate:node01:gpadmin-[INFO]:--------------------------------------------------------------
2.模拟master宕机
在master节点下
# 切换用户
su - gpadmin
# 停止master
pg_ctl stop -s -m fast -D $MASTER_DATA_DIRECTORY
或者
pg_ctl stop -s -m fast -D /data/master/gpseg-1
返回信息：server stopped
查看standby节点状态(master节点执行)
gpstate -f
20190104:14:56:12:262549 gpstate:node01:gpadmin-[INFO]:-Starting gpstate with args: -f
20190104:14:56:12:262549 gpstate:node01:gpadmin-[INFO]:-local Greenplum Version: 'postgres (Greenplum Database) 5.14.0 build commit:13927a46a9b39527fc1af2cbb1571d4fd394ff88'
20190104:14:56:12:262549 gpstate:node01:gpadmin-[CRITICAL]:-gpstate failed. (Reason='could not connect to server: Connection refused
Is the server running on host "localhost" (::1) and accepting
TCP/IP connections on port 5432?
could not connect to server: Connection refused
Is the server running on host "localhost" (127.0.0.1) and accepting
TCP/IP connections on port 5432?
') exiting...集群进入异常状态，无法获取信息
3.激活standby
（1）修改两个文件
su - dbadmin
vi .bashrc
vi .bash_profile
source /usr/local/greenplum-db/greenplum_path.sh
export MASTER_DATA_DIRECTORY=/data/master/gpseg-1
cat >>/home/gpadmin/.bashrc<<-EOF
export PGPORT=5432
EOF
生效：
source ~/.bashrc
（2）激活standby节点
gpactivatestandby -d $MASTER_DATA_DIRECTORY
Do you want to continue with standby master activation? Yy|Nn (default=N):
Y 输入Y确认切换
再次查看standby节点状态
gpstate -f
20190104:15:05:17:199264 gpstate:node02:gpadmin-[INFO]:-Starting gpstate with args: -f
20190104:15:05:17:199264 gpstate:node02:gpadmin-[INFO]:-local Greenplum Version: 'postgres (Greenplum Database) 5.14.0 build commit:13927a46a9b39527fc1af2cbb1571d4fd394ff88'
20190104:15:05:17:199264 gpstate:node02:gpadmin-[INFO]:-master Greenplum Version: 'PostgreSQL 8.3.23 (Greenplum Database 5.14.0 build commit:13927a46a9b39527fc1af2cbb1571d4fd394ff88) on x86_64-pc-linux-gnu, compiled by GCC gcc (GCC) 6.2.0, 64-bit compiled on Nov 28 2018 08:08:06'
20190104:15:05:17:199264 gpstate:node02:gpadmin-[INFO]:-Obtaining Segment details from master...
20190104:15:05:18:199264 gpstate:node02:gpadmin-[INFO]:-Standby master instance not configured
20190104:15:05:18:199264 gpstate:node02:gpadmin-[INFO]:--------------------------------------------------------------
20190104:15:05:18:199264 gpstate:node02:gpadmin-[INFO]:--pg_stat_replication
20190104:15:05:18:199264 gpstate:node02:gpadmin-[INFO]:--------------------------------------------------------------
20190104:15:05:18:199264 gpstate:node02:gpadmin-[INFO]:-No entries found.
20190104:15:05:18:199264 gpstate:node02:gpadmin-[INFO]:--------------------------------------------------------------集群可以查看状态，无法获取信息备用节点信息，此时集群可以访问；
将原有主节点添加为standby （master节点，gpadmin账户操作）
这里添加的时候，会check master的数据库目录，要把原master节点的的目录删除或者重命名
mv /data/master/gpseg-1 /data/master/gpseg-1.bak
将原有主节点先添加为standby节点(在standby节点)
gpinitstandby -s node01
到此为止主备切换完成