重庆分公司,新征程启航

为企业提供网站建设、域名注册、服务器等服务

怎么在Postgresql中选择关闭模式-创新互联

这篇文章将为大家详细讲解有关怎么在Postgresql中选择关闭模式,文章内容质量较高,因此小编分享给大家做个参考,希望大家阅读完这篇文章后对相关知识有一定的了解。

目前创新互联建站已为近千家的企业提供了网站建设、域名、虚拟空间、绵阳服务器托管、企业网站设计、万荣网站维护等服务,公司将坚持客户导向、应用为本的策略,正道将秉承"和谐、参与、激情"的文化,与客户和合作伙伴齐心协力一起成长,共同发展。

停止数据库的命令:

pg_ctl stop -D $PGDATA [-m shutdown-mode]

shutdown-mode有如下几种模式:

1. smart: 等所有的连接中止后,关闭数据库。如果客户端连接不终止, 则无法关闭数据库。

开启一个空会话:

[root@localhost ~]# su - postgres
[postgres@localhost ~]$ psql
psql (9.4.4)
Type "help" for help.
 
postgres=#

用smart关闭数据库:

[postgres@localhost ~]$ pg_ctl stop -D $PGDATA -m smart
waiting for server to shut down............................................................... failed
pg_ctl: server does not shut down
HINT: The "-m fast" option immediately disconnects sessions rather than
waiting for session-initiated disconnection

2. fast: 快速关闭数据库, 断开客户端的连接,让已有的事务回滚,然后正常关闭数据库。

[postgres@localhost ~]$ pg_ctl stop -D $PGDATA -m fast
waiting for server to shut down.... done
server stopped

查看关闭日志:

LOG: received fast shutdown request
LOG: aborting any active transactions
FATAL: terminating connection due to administrator command
LOG: shutting down
LOG: database system is shut down

会话被强制中断,然后关闭数据库。

起一个事务,然后测试关闭:

postgres=# create table t(id int primary key, name varchar(9));
CREATE TABLE
postgres=# begin;
BEGIN
postgres=# insert into t values(1,'a')
postgres-# ;
INSERT 0 1

不提交, 然后用FAST MODE去关闭数据库:

[postgres@localhost ~]$ pg_ctl stop -D $PGDATA -m fast
waiting for server to shut down.... done
server stopped

查看日志:

LOG: received fast shutdown request
LOG: aborting any active transactions
LOG: autovacuum launcher shutting down
FATAL: terminating connection due to administrator command
LOG: shutting down
LOG: database system is shut down

同样是直接中断会话, 而不去管事务有没有提交。

postgres=# select * from t;
id | name
----+------
(0 rows)

没有提交的数据, 在重启之后并不能查到。

3. immediate: 立即关闭数据库,立即停止数据库进程,直接退出,下次启动时会进行实例恢复。

postgres=# insert into t values(1,'a')
;
INSERT 0 1
postgres=# select * from t;
id | name
----+------
1 | a
(1 row)

关闭数据库:

[postgres@localhost ~]$ pg_ctl stop -D $PGDATA -m immediate
waiting for server to shut down.... done
server stopped

查看日志:

LOG: received immediate shutdown request
WARNING: terminating connection because of crash of another server process
DETAIL: The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.
HINT: In a moment you should be able to reconnect to the database and repeat your command.
WARNING: terminating connection because of crash of another server process
DETAIL: The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.
HINT: In a moment you should be able to reconnect to the database and repeat your command.

启动数据库:

[postgres@localhost ~]$ pg_ctl -D /apps/pgsql/pgdata -l 1.log start
server starting

查看日志:

LOG: database system was interrupted; last known up at 2017-04-27 18:56:47 PDT
LOG: database system was not properly shut down; automatic recovery in progress #提示非正常关机,自动开启恢复。
LOG: redo starts at 0/181F910
LOG: record with zero length at 0/181FA90
LOG: redo done at 0/181FA60
LOG: last completed transaction was at log time 2017-04-27 18:59:13.727213-07
LOG: MultiXact member wraparound protections are now enabled
LOG: autovacuum launcher started
LOG: database system is ready to accept connections

查看数据:

[postgres@localhost ~]$ psql
psql (9.4.4)
Type "help" for help.
 
postgres=# select * from t;
id | name
----+------
1 | a
(1 row)

提交的数据已通过实例恢复。

小结:

对比以上三种关库模式:

smart最为安全,但最慢, 需要将所有连接都断开后,才会关库,默认关库模式。

fast强制中断会话,而不管有操作有没有提交,在做系统维护(系统维护时一般应用都正常关闭了,或者不再会有事务操作。)时,需要这种模式来关闭数据库。

immediate最暴力的方式,不管数据有没有落盘(POSGRE是遵循WAL机制),就直接关掉, 待启动时进行实例恢复, 如果在关库前有大量的事务没有写入磁盘, 那这个恢复过程可能会非常的漫长。

补充:postgresql 异步 stream replication 环境关闭 master 的验证

os: ubuntu 16.04

db: postgresql 9.6.8

验证在异步 stream replication环境下,主动关闭master时,数据是否有丢失,能丢失多少。

版本

# lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 16.04.5 LTS
Release: 16.04
Codename: xenial
# su - postgres -c "psql -c \"select version();\""
                 version                 
----------------------------------------------------------------------------------------------------------------------------------------------
 PostgreSQL 9.6.8 on x86_64-pc-linux-gnu (Ubuntu 9.6.8-1.pgdg16.04+1), compiled by gcc (Ubuntu 5.4.0-6ubuntu1~16.04.9) 5.4.0 20160609, 64-bit
(1 row)

用 pgbench 模拟数据库的大量数据操作

postgres=# create database pgbenchdb;
# su - postgres
$ pgbench -i -s 20 pgbenchdb
$ pgbench -r -j2 -c4 -T300 pgbenchdb

关闭 master

# su - postgres
$ /usr/lib/postgresql/9.6/bin/pg_ctl stop -m fast -D "/data/pg9.6/main"

提升 slave

# su - postgres
$ /usr/lib/postgresql/9.6/bin/pg_ctl promote -D "/data/pg9.6/main"

查看 old master 的 xlog location

$ /usr/lib/postgresql/9.6/bin/pg_xlogdump 000000010000000000000016
rmgr: XLOG  len (rec/tot): 106/ 106, tx:   0, lsn: 0/16000028, prev 0/152C9A10, desc: CHECKPOINT_SHUTDOWN redo 0/16000028; tli 1; prev tli 1; fpw true; xid 0:118746; oid 16432; multi 1; offset 0; oldest xid 543 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 550/118745; oldest running xid 0; shutdown
pg_xlogdump: FATAL: error in WAL record at 0/16000028: invalid record length at 0/16000098: wanted 24, got 0

可以看到 lsn: 0/16000028, prev 0/152C9A10, desc: CHECKPOINT_SHUTDOWN redo 0/16000028;

查看 new master 的 .history文件

$ ls -lt|more
total 360456
-rw------- 1 postgres postgres 16777216 Nov 30 10:32 000000020000000000000016
drwx------ 2 postgres postgres  4096 Nov 30 10:16 archive_status
-rw------- 1 postgres postgres  42 Nov 30 10:16 00000002.history
-rw------- 1 postgres postgres 16777216 Nov 30 10:16 000000010000000000000016.partial
-rw------- 1 postgres postgres 16777216 Nov 30 10:16 000000010000000000000015
-rw------- 1 postgres postgres 16777216 Nov 30 10:16 000000010000000000000014
-rw------- 1 postgres postgres 16777216 Nov 30 10:05 000000010000000000000013
$ cat 00000002.history 
1 0/16000098 no recovery target specified
$ /usr/lib/postgresql/9.6/bin/pg_xlogdump 000000010000000000000016
rmgr: XLOG  len (rec/tot): 106/ 106, tx:   0, lsn: 0/16000028, prev 0/152C9A10, desc: CHECKPOINT_SHUTDOWN redo 0/16000028; tli 1; prev tli 1; fpw true; xid 0:118746; oid 16432; multi 1; offset 0; oldest xid 543 in DB 1; oldest multi 1 in DB 1; oldest/newest commit timestamp xid: 550/118745; oldest running xid 0; shutdown
rmgr: XLOG  len (rec/tot):  42/ 42, tx:   0, lsn: 0/16000098, prev 0/16000028, desc: END_OF_RECOVERY tli 2; prev tli 1; time 2018-11-30 10:16:57.249408 CST
rmgr: Standby  len (rec/tot):  50/ 50, tx:   0, lsn: 0/160000C8, prev 0/16000098, desc: RUNNING_XACTS nextXid 118746 latestCompletedXid 118745 oldestRunningXid 118746
rmgr: XLOG  len (rec/tot):  51/ 312, tx:   0, lsn: 0/16000100, prev 0/160000C8, desc: FPI_FOR_HINT , blkref #0: rel 1664/0/1260 blk 0 FPW
rmgr: Standby  len (rec/tot):  50/ 50, tx:   0, lsn: 0/16000238, prev 0/16000100, desc: RUNNING_XACTS nextXid 118746 latestCompletedXid 118745 oldestRunningXid 118746
pg_xlogdump: FATAL: error in WAL record at 0/16000238: invalid record length at 0/16000270: wanted 24, got 0

可以看到关键记录

lsn: 0/16000028, prev 0/152C9A10, desc: CHECKPOINT_SHUTDOWN redo 0/16000028
lsn: 0/16000098, prev 0/16000028, desc: END_OF_RECOVERY

而 END_OF_RECOVERY 对应的 lsn 为 0/16000098,和 00000002.history 时间线文件的内容完全一致。

所以在异步 stream replication 环境下,主动关闭master时,会将最后一条记录(CHECKPOINT_SHUTDOWN)发送给slave,不会造成数据的丢失。

而 synchronous_commit = on 保证事务有两份持久化的落盘数据。

分析 pg_log 日志

old master 上的最后几条日志

2018-11-30 10:16:40.986 CST,"postgres","pgbenchdb",7559,"[local]",5c009d79.1d87,4,"UPDATE waiting",2018-11-30 10:16:25 CST,,0,LOG,00000,"disconnection: session time: 0:00:15.723 user=postgres database=pgbenchdb host=[local]",,,,,,,,,"pgbench"
2018-11-30 10:16:40.993 CST,"postgres","pgbenchdb",7558,"[local]",5c009d79.1d86,3,"idle",2018-11-30 10:16:25 CST,4/0,0,FATAL,57P01,"terminating connection due to administrator command",,,,,,,,,"pgbench"
2018-11-30 10:16:40.994 CST,"postgres","pgbenchdb",7560,"[local]",5c009d79.1d88,3,"idle",2018-11-30 10:16:25 CST,5/0,0,FATAL,57P01,"terminating connection due to administrator command",,,,,,,,,"pgbench"
2018-11-30 10:16:40.994 CST,"postgres","pgbenchdb",7558,"[local]",5c009d79.1d86,4,"idle",2018-11-30 10:16:25 CST,,0,LOG,00000,"disconnection: session time: 0:00:15.729 user=postgres database=pgbenchdb host=[local]",,,,,,,,,"pgbench"
2018-11-30 10:16:40.994 CST,"postgres","pgbenchdb",7560,"[local]",5c009d79.1d88,4,"idle",2018-11-30 10:16:25 CST,,0,LOG,00000,"disconnection: session time: 0:00:15.725 user=postgres database=pgbenchdb host=[local]",,,,,,,,,"pgbench"
2018-11-30 10:16:40.999 CST,"postgres","pgbenchdb",7561,"[local]",5c009d79.1d89,3,"idle",2018-11-30 10:16:25 CST,6/0,0,FATAL,57P01,"terminating connection due to administrator command",,,,,,,,,"pgbench"
2018-11-30 10:16:41.001 CST,"postgres","pgbenchdb",7561,"[local]",5c009d79.1d89,4,"idle",2018-11-30 10:16:25 CST,,0,LOG,00000,"disconnection: session time: 0:00:15.731 user=postgres database=pgbenchdb host=[local]",,,,,,,,,"pgbench"
2018-11-30 10:16:41.010 CST,,,7156,,5c009735.1bf4,7,,2018-11-30 09:49:41 CST,,0,LOG,00000,"shutting down",,,,,,,,,""
2018-11-30 10:16:41.209 CST,,,7156,,5c009735.1bf4,8,,2018-11-30 09:49:41 CST,,0,LOG,00000,"checkpoint starting: shutdown immediate",,,,,,,,,""
2018-11-30 10:16:47.623 CST,,,7156,,5c009735.1bf4,9,,2018-11-30 09:49:41 CST,,0,LOG,00000,"checkpoint complete: wrote 29357 buffers (89.6%); 0 transaction log file(s) added, 0 removed, 0 recycled; write=5.931 s, sync=0.399 s, total=6.418 s; sync files=53, longest=0.179 s, average=0.007 s; distance=311294 kB, estimate=311294 kB",,,,,,,,,""
2018-11-30 10:16:47.683 CST,"repl","",7227,"192.168.56.90:52556",5c009795.1c3b,3,"streaming 0/16000098",2018-11-30 09:51:17 CST,,0,LOG,00000,"disconnection: session time: 0:25:30.149 user=repl database= host=192.168.56.90 port=52556",,,,,,,,,"walreceiver"
2018-11-30 10:16:47.730 CST,,,7153,,5c009735.1bf1,5,,2018-11-30 09:49:41 CST,,0,LOG,00000,"database system is shut down",,,,,,,,,""

注意倒数第二条信息 streaming 0/16000098 ,说明当时的master关闭时,已经和salve沟通过,确认已经接收到 END_OF_RECOVERY 之前所有的数据了。

old slave 日志

2018-11-30 10:16:47.660 CST,,,7256,,5c009795.1c58,2,,2018-11-30 09:51:17 CST,,0,LOG,00000,"replication terminated by primary server","End of WAL reached on timeline 1 at 0/16000098.",,,,,,,,""
2018-11-30 10:16:47.660 CST,,,7256,,5c009795.1c58,3,,2018-11-30 09:51:17 CST,,0,FATAL,XX000,"could not send end-of-streaming message to primary: no COPY in progress
",,,,,,,,,""
2018-11-30 10:16:47.660 CST,,,7255,,5c009795.1c57,5,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"invalid record length at 0/16000098: wanted 24, got 0",,,,,,,,,""
2018-11-30 10:16:47.854 CST,,,7443,,5c009d8f.1d13,1,,2018-11-30 10:16:47 CST,,0,FATAL,XX000,"could not connect to the primary server: could not connect to server: Connection refused
 Is the server running on host ""192.168.56.119"" and accepting
 TCP/IP connections on port 5432?
",,,,,,,,,""
2018-11-30 10:16:52.668 CST,,,7444,,5c009d94.1d14,1,,2018-11-30 10:16:52 CST,,0,FATAL,XX000,"could not connect to the primary server: could not connect to server: Connection refused
 Is the server running on host ""192.168.56.119"" and accepting
 TCP/IP connections on port 5432?
",,,,,,,,,""
2018-11-30 10:16:56.875 CST,,,7255,,5c009795.1c57,6,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"received promote request",,,,,,,,,""
2018-11-30 10:16:56.875 CST,,,7255,,5c009795.1c57,7,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"redo done at 0/16000028",,,,,,,,,""
2018-11-30 10:16:56.875 CST,,,7255,,5c009795.1c57,8,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"last completed transaction was at log time 2018-11-30 10:16:40.986869+08",,,,,,,,,""
2018-11-30 10:16:56.888 CST,,,7255,,5c009795.1c57,9,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"selected new timeline ID: 2",,,,,,,,,""
2018-11-30 10:16:57.166 CST,,,7255,,5c009795.1c57,10,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"archive recovery complete",,,,,,,,,""
2018-11-30 10:16:57.267 CST,,,7255,,5c009795.1c57,11,,2018-11-30 09:51:17 CST,1/0,0,LOG,00000,"MultiXact member wraparound protections are now enabled",,,,,,,,,""
2018-11-30 10:16:57.267 CST,,,7257,,5c009795.1c59,1,,2018-11-30 09:51:17 CST,,0,LOG,00000,"checkpoint starting: force",,,,,,,,,""
2018-11-30 10:16:57.275 CST,,,7253,,5c009795.1c55,3,,2018-11-30 09:51:17 CST,,0,LOG,00000,"database system is ready to accept connections",,,,,,,,,""
2018-11-30 10:16:57.276 CST,,,7447,,5c009d99.1d17,1,,2018-11-30 10:16:57 CST,,0,LOG,00000,"autovacuum launcher started",,,,,,,,,""

信息也是相当的清晰。

wal_retrieve_retry_interval = 5s 控制 salve 到 master 失败时,再次重试的等待时间。

关于怎么在Postgresql中选择关闭模式就分享到这里了,希望以上内容可以对大家有一定的帮助,可以学到更多知识。如果觉得文章不错,可以把它分享出去让更多的人看到。


网站题目:怎么在Postgresql中选择关闭模式-创新互联
文章分享:http://cqcxhl.com/article/ddcphs.html

其他资讯

在线咨询
服务热线
服务热线:028-86922220
TOP