postgresql表死锁问题的排查方式

时间:2022-06-01 19:21:50

1.查询激活的执行中的sql,查看有哪些更新update的sql。

?
1
2
3
select *
from pg_stat_activity
where state = 'active';

2. 查询表中存在的锁

?
1
2
3
4
select a.locktype, a.database, a.pid, a.mode, a.relation, b.relname
from pg_locks a
join pg_class b on a.relation = b.oid
where lower(b.relname) = 'h5_game';

3. 杀掉死锁进程

?
1
2
3
4
5
6
7
8
9
select pg_terminate_backend(pid)
from pg_stat_activity
where state = 'active'
and pid != pg_backend_pid()
--and pid = 14172
and pid in (select a.pid
from pg_locks a
join pg_class b on a.relation = b.oid
where lower(b.relname) = 'news_content')

锁模式

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
/* NoLock is not a lock mode, but a flag value meaning "don't get a lock" */
#define NoLock                 0
 
#define AccessShareLock         1        /* SELECT */
#define RowShareLock          2        /* SELECT FOR UPDATE/FOR SHARE */
#define RowExclusiveLock        3        /* INSERT, UPDATE, DELETE */
#define ShareUpdateExclusiveLock 4       /* VACUUM (non-FULL),ANALYZE, CREATE
                                         * INDEX CONCURRENTLY */
#define ShareLock                5        /* CREATE INDEX (WITHOUT CONCURRENTLY) */
#define ShareRowExclusiveLock  6        /* like EXCLUSIVE MODE, but allows ROW
                                         * SHARE */
#define ExclusiveLock          7        /* blocks ROW SHARE/SELECT...FOR
                                         * UPDATE */
#define AccessExclusiveLock       8        /* ALTER TABLE, DROP TABLE, VACUUM
                                         * FULL, and unqualified LOCK TABLE */

补充:Postgresql死锁的处理

背景:

对表进行所有操作都卡住,原因可能是更新表时导致这个表死锁了,开始进行排查

解决一:查询pg_stat_activity有没有记录

 

pg版本10.2

?
1
2
3
select pid,query,* from pg_stat_activity where datname='死锁的数据库' and wait_event_type = 'Lock';
select pg_cancel_backend('死锁那条数据的pid值');##只能杀死select 语句, 对其他语句不生效
pg_terminate_backend('死锁那条数据的pid值');#select,drop等各种操作

执行后发现select和delete表时正常执行,但truncate和drop表时会一直运行,也不报错。

“drop table” 和 “truncate table” 需要申请排它锁"ACCESS EXCLUSIVE", 执行这个命令卡住时,说明此时这张表上还有操作正在进行,比如查询等,

那么只有等待这个查询操作完成,“drop table” 或"truncate table"或者增加字段的SQL才能获取这张表上的 "ACCESS EXCLUSIVE"锁,操作才能进行下去。

解决二:查询pg_locks是否有这个对象的锁

 

?
1
2
3
select oid,relname from pg_class where relname='table name';
select locktype,pid,relation,mode,granted,* from pg_locks where relation= '上面查询出来的oid';
select pg_terminate_backend('进程ID');

问题解决!!!

坑:一开始不知道pg_cancel_backend(‘死锁那条数据的pid值');##只能杀死select 语句, 对其他语句不生效,杀了进程查询发现还存在,反复杀反复存在,换了pg_terminate_backend(‘进程ID')问题就解决了。

以上为个人经验,希望能给大家一个参考,也希望大家多多支持服务器之家。如有错误或未考虑完全的地方,望不吝赐教。

原文链接:https://blog.csdn.net/fsstyle/article/details/87917720