问题描述
我有 3 个节点 Postgres 集群由 patroni 管理,每当我的主节点出现故障并作为副本加入时,旧主节点就会面临以下错误:
2021-06-25T00:16:29.856133+00:00 host0 postgres_0[14131]: [7-2] #011DETAIL: This server's history forked from timeline 1 at 0/5208600.
2021-06-25T00:16:29.862228+00:00 host0 postgres_0[112]: [4855-1] pid=112,session=60d4c1b1.70,line=4850,sqlstate=00000,user_app=,user=,db=,client=,txId=0 LOG: new timeline 2 forked off current database system timeline 1 before current recovery point 0/60000A0
2021-06-25T00:16:34.857325+00:00 host0 postgres_0[14141]: [7-1] pid=14141,session=60d52062.373d,line=1,sqlstate=XX000,txId=0 FATAL: Could not start WAL streaming: ERROR: requested starting point 0/6000000 on timeline 1 is not in this server's history
这是我使用的配置:
"hot_standby": "on","wal_log_hints": "on","restore_command": "cp /bp2/wal/psql/wal_archive/%f %p","archive_mode": "on","archive_command": "mkdir -p /bp2/wal/psql/wal_archive && test ! -f /bp2/wal/psql/wal_archive/%f && cp %p /bp2/wal/psql/wal_archive/%f","remove_data_directory_on_diverged_timelines":"true","remove_data_directory_on_rewind_failure": "true","use_pg_rewind": "true","recovery_target_timeline": "latest"
已经尝试将所有节点的 WAL 日志存储到共享目录中,然后从那里恢复。
但是错误还是一样。
解决方法
我认为您需要做的是,在您的新副本中手动使用 pg_rewind
命令。发生的情况是,当时间线从 1 变为 2 时,旧主控与新主控处于不同的 wal 位置。
pg_rewind --source-server "user=<user> password=<user_password> host=<dns_or_ip_address> port=<server_port>" --target-pgdata <data-directory>