postgresql 日志配置

- 日理万妓 2022-08-04 00:44 402阅读 0赞

PostgreSQL有3种日志，分别是pg\_log（数据库运行日志）、pg\_xlog（WAL 日志，即重做日志）、pg\_clog（事务提交日志，记录的是事务的元数据）  
  
  
pg\_log默认是关闭的，需要设置参数启用此日志。pg\_xlog和pg\_clog都是强制打开的，无法关闭。  
  
  
  
  
  
  
1.启用pg\_log并配置日志参数  
log\_destination = 'csvlog'  
logging\_collector = on  
log\_directory = 'pg\_log'  
log\_filename = 'postgresql-%Y-%m-%d\_%H%M%S.log'  
log\_rotation\_age = 1d  
log\_rotation\_size = 100MB  
log\_min\_messages = info  
\# 记录执行慢的SQL  
log\_min\_duration\_statement = 60  
log\_checkpoints = on  
log\_connections = on  
log\_disconnections = on  
log\_duration = on  
log\_line\_prefix = '%m'  
\# 监控数据库中长时间的锁  
log\_lock\_waits = on  
\# 记录DDL操作  
log\_statement = 'ddl'  
  
  
2.重启PostgreSQL即可在$PGDATA/pg\_log/ 下看到新生成的日志。

pg\_ctl restart -m fast

**--日志文件目录   **  
   日志的目录可以通过参数  log\_directory 来设置，下面是我的参数设置。  
   log\_directory = '/var/applog/pg\_log'    
      
\[postgres@pg\_log\]$ ll /var/applog/pg\_log      
\-rw------- 1 postgres postgres 4.8M Mar 14 23:57 postgresql-2011-03-14\_000000.csv  
\-rw------- 1 postgres postgres    0 Mar 14 00:00 postgresql-2011-03-14\_000000.log  
\-rw------- 1 postgres postgres 294K Mar 15 15:10 postgresql-2011-03-15\_000000.csv  
\-rw------- 1 postgres postgres    0 Mar 15 00:00 postgresql-2011-03-15\_000000.log

**--CSV日志文件内容**  
2011-03-15 00:07:03.513 CST,"wapportal","wapportal\_216",4137,"172.16.3.43:59356",4d7e361f.1029,3,"idle",2011-03-14 23:37:03 CST,,0,LOG,00000,"disconnection: session time: 0:30:00.086 user=wapportal database=wapportal\_216 host=172.16.3.43 port=59356",,,,,,,,,""  
2011-03-15 00:07:03.514 CST,,,5173,"",4d7e3d27.1435,1,"",2011-03-15 00:07:03 CST,,0,LOG,00000,"connection received: host=172.16.3.43 port=51135",,,,,,,,,""

上面两条是 postgresql-2011-03-15\_000000.csv 日志文件的部分内容 ,由于日志文件的可读性  
 较差，于是可以通过下面方法将CSV日志导入到数据库表里。详细如下

**将CSV日志导入数据库表里   **  
**1--调整参数**  
log\_destination = 'csvlog'  
logging\_collector = on

这两个参数修改后，PG SERVER 需要重启。

**2--创建日志记录表**  
CREATE TABLE postgres\_log  
(  
  log\_time timestamp(3) with time zone,  
  user\_name text,  
  database\_name text,  
  process\_id integer,  
  connection\_from text,  
  session\_id text,  
  session\_line\_num bigint,  
  command\_tag text,  
  session\_start\_time timestamp with time zone,  
  virtual\_transaction\_id text,  
  transaction\_id bigint,  
  error\_severity text,  
  sql\_state\_code text,  
  message text,  
  detail text,  
  hint text,  
  internal\_query text,  
  internal\_query\_pos integer,  
  context text,  
  query text,  
  query\_pos integer,  
  location text,  
  application\_name text,  
  PRIMARY KEY (session\_id, session\_line\_num)  
);

NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "postgres\_log\_pkey" for table "postgres\_log"  
CREATE TABLE;

备注：创建日志表 postgres\_log 用来保存 CSV日志数据。

**3--导入操作系统 csv 日志到表 postgres\_log 表**  
skytf=\# copy skytf.postgres\_log from '/var/applog/pg\_log/postgresql-2011-03-14\_000000.csv' with csv;  
COPY 26031

skytf=\# copy skytf.postgres\_log from '/var/applog/pg\_log/postgresql-2011-03-15\_000000.csv' with csv;  
COPY 1297

备注：文件形式导入导出数据需要以超级用户 postgres 连接到目标库。

**4--常用日志分析sql**  
skytf=\# select min(log\_time),max(log\_time) from skytf.postgres\_log;  
            min             |            max               
\----------------------------+----------------------------  
 2011-03-14 14:04:07.275+08 | 2011-03-16 05:04:34.427+08  
(1 row)

skytf=> select log\_time,database\_name,user\_name,application\_name,message from postgres\_log where message like '%duration%';  
          log\_time          | database\_name | user\_name | application\_name |                                                    mess  
age                                                      
\----------------------------+---------------+-----------+------------------+--------------------------------------------------------  
\-------------------------------------------------------  
 2011-03-15 00:23:38.957+08 | db\_lbs        | lbs       |                  | duration: 1297.440 ms  execute <unnamed>:    SELECT cit  
yname,province,the\_geom as the\_geom FROM china\_city     
.......  
      
       为了显示方便，上面只取一条记录。