PG14归档失败解决办法archiver failed on wal_lsn

问题描述

昨晚Repmgr+PG14主备主库因wal日志撑爆磁盘，删除主库过期wal文件重做备库后上午进行主备状态巡查，主库向备库发送wal文件正常，但是查主库状态时发现显示有1条归档失败的记录。
postgres: archiver failed on 000000010000006F00000086

• 主库：

walsender repmgr 172.28.32.23(36122) streaming 72/1BAC3A10" walsender正常
archiver failed on 000000010000006F00000086" 归档失败

• 备库：

walreceiver streaming 77/9EB6A198" “” “” " walreceiver正常

--查主库数据库状态
[root@pgmaster ~]# systemctl status postgres
● postgres.service - PostgreSQL database server
Loaded: loaded (/usr/lib/systemd/system/postgres.service; enabled; vendor preset: disabled)
Active: active (running) since Thu 2023-10-12 22:04:08 CST; 13h ago
Process: 3710968 ExecStart=/server/data/pgdb/pgsql/bin/pg_ctl start -D $PGDATA (code=exited, status=0/SUCCESS)
Main PID: 3710970 (postgres)
Tasks: 53 (limit: 201967)
Memory: 19.0G
CGroup: /system.slice/postgres.service
├─ 3710970 /server/data/pgdb/pgsql/bin/postgres -D /server/data/pgdb/data
├─ 3710971 "postgres: logger " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710992 "postgres: checkpointer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710993 "postgres: background writer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710994 "postgres: walwriter " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710995 "postgres: archiver failed on 000000010000006F00000086" "" "" "" "" "" "" "" "" ""
├─ 3710996 "postgres: logical replication launcher " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3711001 "postgres: top_portal top_portal 172.28.32.18(41438) idle" "" "" "" "" "" ""
├─ 3711003 "postgres: tj_sjjh dataexchange 172.28.32.28(35406) idle" "" "" "" "" "" "" ""
├─ 3711009 "postgres: repmgr repmgr 172.28.32.22(64096) idle" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3711468 "postgres: top_portal top_portal 172.28.32.18(41720) idle" "" "" "" "" "" ""
├─ 3713807 "postgres: top_portal top_portal 172.28.32.20(44492) idle" "" "" "" "" "" ""
├─ 3723017 "postgres: walsender repmgr 172.28.32.23(36122) streaming 72/1BAC3A10"  #wal 发送正常

--查备库状态
[root@pgslave ~]# systemctl status postgres
● postgres.service - PostgreSQL database server
Loaded: loaded (/usr/lib/systemd/system/postgres.service; enabled; vendor preset: disabled)
Active: active (running) since Fri 2023-10-13 00:12:19 CST; 12h ago
Process: 1931221 ExecStart=/server/data/pgdb/pgsql/bin/pg_ctl start -D $PGDATA (code=exited, status=0/SUCCESS)
Main PID: 1931223 (postgres)
Tasks: 7 (limit: 201967)
Memory: 23.2G
CGroup: /system.slice/postgres.service
├─ 1931223 /server/data/pgdb/pgsql/bin/postgres -D /server/data/pgdb/data
├─ 1931224 "postgres: logger " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 1931225 "postgres: startup recovering 00000001000000770000009E" "" "" "" "" "" "" "" "" ""
├─ 1931226 "postgres: checkpointer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 1931227 "postgres: background writer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 1931230 "postgres: walreceiver streaming 77/9EB6A198" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""   #wal接收
└─ 1931430 "postgres: repmgr repmgr 172.28.32.23(22956) idle" "" "" "" "" "" "" "" "" "" "" "" "" "" ""

Oct 13 00:12:17 pgslave systemd[1]: Starting PostgreSQL database server...
Oct 13 00:12:17 pgslave pg_ctl[1931221]: waiting for server to start....
Oct 13 00:12:17 pgslave pg_ctl[1931223]: 2023-10-13 00:12:17.497 CST [1931223] LOG:  redirecting log output to logging collector process
Oct 13 00:12:17 pgslave pg_ctl[1931223]: 2023-10-13 00:12:17.497 CST [1931223] HINT:  Future log output will appear in directory "log".
Oct 13 00:12:19 pgslave pg_ctl[1931221]: . done
Oct 13 00:12:19 pgslave pg_ctl[1931221]: server started
Oct 13 00:12:19 pgslave systemd[1]: Started PostgreSQL database server.

问题分析

1.查看数据库日志

2.查看归档配置参数

参数配置正确，归档目录权限也正确

postgres=# show archive_command;
                      archive_command                      
-----------------------------------------------------------
 /usr/bin/lz4 -q -z %p /server/data/pgdb/pg_archive/%f.lz4
(1 row)

postgres=# show archive_mode;
 archive_mode 
--------------
 on
(1 row)

--查看归档目录的权限
[postgres@pgmaster ~]$ ls -ld /server/data/pgdb/pg_archive
drwxr-x--- 2 postgres postgres 4214784 Oct 13 13:14 /server/data/pgdb/pg_archive

3.手动切日志

手工归档成功，但是未解决，查看状态依然时卡住归档失败的那条wal记录那里

--手工归档
top_portal=# select pg_switch_wal();
 pg_switch_wal 
---------------
 72/51C4CFD8
(1 row)

--查主库数据库状态
[root@pgmaster ~]# systemctl status postgres
● postgres.service - PostgreSQL database server
Loaded: loaded (/usr/lib/systemd/system/postgres.service; enabled; vendor preset: disabled)
Active: active (running) since Thu 2023-10-12 22:04:08 CST; 13h ago
Process: 3710968 ExecStart=/server/data/pgdb/pgsql/bin/pg_ctl start -D $PGDATA (code=exited, status=0/SUCCESS)
Main PID: 3710970 (postgres)
Tasks: 53 (limit: 201967)
Memory: 19.0G
CGroup: /system.slice/postgres.service
├─ 3710970 /server/data/pgdb/pgsql/bin/postgres -D /server/data/pgdb/data
├─ 3710971 "postgres: logger " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710992 "postgres: checkpointer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710993 "postgres: background writer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710994 "postgres: walwriter " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3710995 "postgres: archiver failed on 000000010000006F00000086" "" "" "" "" "" "" "" "" ""
├─ 3710996 "postgres: logical replication launcher " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3711001 "postgres: top_portal top_portal 172.28.32.18(41438) idle" "" "" "" "" "" ""
├─ 3711003 "postgres: tj_sjjh dataexchange 172.28.32.28(35406) idle" "" "" "" "" "" "" ""
├─ 3711009 "postgres: repmgr repmgr 172.28.32.22(64096) idle" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
├─ 3711468 "postgres: top_portal top_portal 172.28.32.18(41720) idle" "" "" "" "" "" ""
├─ 3713807 "postgres: top_portal top_portal 172.28.32.20(44492) idle" "" "" "" "" "" ""
├─ 3723017 "postgres: walsender repmgr 172.28.32.23(36122) streaming 72/1BAC3A10"  #wal 发送正常

--查当前wal_lsn
top_portal=# select pg_current_wal_lsn();
 pg_current_wal_lsn 
--------------------
 72/52638F10
(1 row)

--查当前wal_lsn对应的wal文件
top_portal=# select pg_walfile_name(pg_current_wal_lsn());
     pg_walfile_name      
--------------------------
 000000010000007200000052
(1 row)

--查当前最新检查点,最新检查点之前的wal文件均可以删除
[postgres@pgmaster ~]$ pg_controldata $PGDATA
pg_control version number:            1300
Catalog version number:               202107181
Database system identifier:           7268852449124462799
Database cluster state:               in production
pg_control last modified:             Fri 13 Oct 2023 10:07:35 AM CST  
Latest checkpoint location:           71/CDD2FF28
Latest checkpoint's REDO location:    71/CDD28F18
Latest checkpoint's REDO WAL file:    0000000100000071000000CD

--查报错中的wal文件
[postgres@pgmaster pg_wal]$ ls -l 000000010000006F00000086
-rw------- 1 postgres postgres 16777216 Oct 12 21:12 000000010000006F00000086
[postgres@pgmaster pg_wal]$ find /server/data/pgdb/pg_archive -name 000000010000006F00000086*
ls: cannot access '000000010000006F00000086': No such file or directory
[postgres@pgmaster pg_wal]$ find /server -name 000000010000006F00000086*
-rw------- 1 postgres postgres 16777216 Oct 12 21:12 000000010000006F00000086

4.检查$PGDATA/pg_wal/archive_status/目录下文件

[postgres@pgmaster ~]$ cd /server/data/pgdb/data/pg_wal/archive_status/
[postgres@pgmaster archive_status]$ ls -l *.ready
ls: cannot access '*.ready': No such file or directory

说明不存在需要归档但没归档的文件

该目录下，ready说明是需要归档但是没归档的，done是归档完成了的

解决办法

1.将归档失败的wal文件备份到/home/postgres目录下(生产环境如果磁盘空间允许切记不要rm删除，mv备份到目标位置)
2.手工归档select pg_switch_wal();
3.再次查看主备库状态

--1.将归档失败的wal文件备份到/home/postgres目录下
[postgres@pgmaster pg_wal]$ mv 000000010000006F00000086 /home/postgres/000000010000006F00000086
[postgres@pgmaster pg_wal]$ ls -l /home/postgres/000000010000006F00000086
-rw------- 1 postgres postgres 16777216 Oct 12 21:12 /home/postgres/000000010000006F00000086

--2.手工归档
postgres=# select pg_switch_wal();
 pg_switch_wal 
---------------
 73/7EF502E0
(1 row)

--3.再次查看主库状态显示正常
[root@pgmaster data]# systemctl status postgres
● postgres.service - PostgreSQL database server
     Loaded: loaded (/usr/lib/systemd/system/postgres.service; enabled; vendor preset: disabled)
     Active: active (running) since Thu 2023-10-12 22:04:08 CST; 13h ago
    Process: 3710968 ExecStart=/server/data/pgdb/pgsql/bin/pg_ctl start -D $PGDATA (code=exited, status=0/SUCCESS)
   Main PID: 3710970 (postgres)
      Tasks: 50 (limit: 201967)
     Memory: 26.6G
     CGroup: /system.slice/postgres.service
             ├─ 3710970 /server/data/pgdb/pgsql/bin/postgres -D /server/data/pgdb/data
             ├─ 3710971 "postgres: logger " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─ 3710992 "postgres: checkpointer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─ 3710993 "postgres: background writer " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─ 3710994 "postgres: walwriter " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─ 3710995 "postgres: archiver archiving 000000010000007100000035" "" "" "" "" "" "" "" "" ""
             ├─ 3710996 "postgres: logical replication launcher " "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─ 3711001 "postgres: top_portal top_portal 172.28.32.18(41438) idle" "" "" "" "" "" ""
             ├─ 3711003 "postgres: tj_sjjh dataexchange 172.28.32.28(35406) idle" "" "" "" "" "" "" ""
             ├─ 3711009 "postgres: repmgr repmgr 172.28.32.22(64096) idle" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─ 3711468 "postgres: top_portal top_portal 172.28.32.18(41720) idle" "" "" "" "" "" ""
             ├─ 3713807 "postgres: top_portal top_portal 172.28.32.20(44492) idle" "" "" "" "" "" ""
             ├─ 3723017 "postgres: walsender repmgr 172.28.32.23(36122) streaming 73/7F000BD0"

补充：若$PGDATA/pg_wal/archive_status/目录下存在大量的*.ready文件

可能的原因分析：如果数据库是突然断电，那么可能arvchive命令没有完全完成，归档目录会存在不完整的文件名称，重启数据库后，会出现归档失败的情况，这个时候，需要去归档目录删除相关归档失败文件，那么归档就会重新归档。
还未遇到该场景的问题，暂未实验。
参考链接