patroni啟動，提示waiting for leader to bootstrap

patroni集群第一次初始化失敗，再次使用原來的配置嘗試初始化，查看信息，不同節點都顯示同一個信息：waiting for leader to bootstrap

#  systemctl status  patroni.service
● patroni.service - PostgreSQL high-availability manager
   Loaded: loaded (/usr/lib/systemd/system/patroni.service; disabled; vendor preset: disabled)
   Active: active (running) since Sat 2020-02-22 17:41:50 CST; 11s ago
 Main PID: 3104 (python3.6)
    Tasks: 5
   CGroup: /system.slice/patroni.service
           └─3104 python3.6 /opt/app/patroni/bin/patroni /opt/app/patroni/etc/postgresql.yml

Feb 22 17:41:50 docker02 systemd[1]: Started PostgreSQL high-availability manager.
Feb 22 17:41:50 docker02 systemd[1]: Starting PostgreSQL high-availability manager...
Feb 22 17:41:50 docker02 patroni[3104]: 2020-02-22 17:41:50,916 INFO: Selected new etcd server http://11.11.11.250:2379
Feb 22 17:41:50 docker02 patroni[3104]: 2020-02-22 17:41:50,928 INFO: No PostgreSQL configuration items changed, nothing to reload.
Feb 22 17:41:50 docker02 patroni[3104]: 2020-02-22 17:41:50,935 INFO: Lock owner: None; I am pg01
Feb 22 17:41:50 docker02 patroni[3104]: 2020-02-22 17:41:50,937 INFO: waiting for leader to bootstrap
Feb 22 17:42:00 docker02 patroni[3104]: 2020-02-22 17:42:00,934 INFO: Lock owner: None; I am pg01
Feb 22 17:42:00 docker02 patroni[3104]: 2020-02-22 17:42:00,939 INFO: waiting for leader to bootstrap

　　

原因：
etcd保留了集群的初始化信息（鍵：/service/$CLUSTER_NAME/initialize）。該鍵存在后，patroni就不會再次執行initdb。而是嘗試再次引導已經存在的節點。

 

解決方法：
1.使用patronictl手動移除etcd中的/service/$CLUSTER_NAME/initialize
2.設置一個新的cluster name后從新引導

 

以下是手動移除的過程：

# patronictl -c /opt/app/patroni/etc/postgresql.yml  list
+---------+--------+--------------+------+---------+----+-----------+
| Cluster | Member |     Host     | Role |  State  | TL | Lag in MB |
+---------+--------+--------------+------+---------+----+-----------+
|  batman |  pg01  | 11.11.11.111 |      | stopped |    |   unknown |
+---------+--------+--------------+------+---------+----+-----------+
# patronictl -c /opt/app/patroni/etc/postgresql.yml  remove batman
+---------+--------+--------------+------+---------+----+-----------+
| Cluster | Member |     Host     | Role |  State  | TL | Lag in MB |
+---------+--------+--------------+------+---------+----+-----------+
|  batman |  pg01  | 11.11.11.111 |      | stopped |    |   unknown |
+---------+--------+--------------+------+---------+----+-----------+
Please confirm the cluster name to remove: batman
You are about to remove all information in DCS for batman, please type: "Yes I am aware": Yes I am aware