本節介紹了與日志相關的配置參數log_XXX。

概覽

在新initdb的數據庫上查詢pg_settings，可查詢PostgreSQL中與log的參數包括：

[test@localhost ~]$ psql
Expanded display is used automatically.
psql (12.1)
Type "help" for help.
[local:/var/run/test]:5000 test@testdb=# select category,name,setting from pg_settings where name like 'log%' order by category;
After canonicalize_qual()
   {OPEXPR 
   :opno 1209 
   :opfuncid 850 
   :opresulttype 16 
   :opretset false 
   :opcollid 0 
   :inputcollid 100 
   :args (
      {VAR 
      :varno 4 
      :varattno 1 
      :vartype 25 
      :vartypmod -1 
      :varcollid 100 
      :varlevelsup 0 
      :varnoold 4 
      :varoattno 1 
      :location -1
      }
      {CONST 
      :consttype 25 
      :consttypmod -1 
      :constcollid 100 
      :constlen -1 
      :constbyval false 
      :constisnull false 
      :location 62 
      :constvalue 8 [ 32 0 0 0 108 111 103 37 ]
      }
   )
   :location 57
   }
RELOPTINFO (a): rows=5 width=96
  baserestrictinfo: a.name ~~ log%
  path list:
  FunctionScan(a) rows=5 cost=0.00..12.50
  cheapest parameterized paths:
  FunctionScan(a) rows=5 cost=0.00..12.50
  cheapest startup path:
  FunctionScan(a) rows=5 cost=0.00..12.50
  cheapest total path:
  FunctionScan(a) rows=5 cost=0.00..12.50
               category               |            name             |            setting             
--------------------------------------+-----------------------------+--------------------------------
 Reporting and Logging / What to Log  | log_lock_waits              | off
 Reporting and Logging / What to Log  | log_checkpoints             | off
 Reporting and Logging / What to Log  | log_connections             | off
 Reporting and Logging / What to Log  | log_timezone                | PRC
 Reporting and Logging / What to Log  | log_temp_files              | -1
 Reporting and Logging / What to Log  | log_disconnections          | off
 Reporting and Logging / What to Log  | log_duration                | off
 Reporting and Logging / What to Log  | log_error_verbosity         | default
 Reporting and Logging / What to Log  | log_statement               | none
 Reporting and Logging / What to Log  | log_replication_commands    | off
 Reporting and Logging / What to Log  | log_autovacuum_min_duration | -1
 Reporting and Logging / What to Log  | log_hostname                | off
 Reporting and Logging / What to Log  | log_line_prefix             | %m [%p] 
 Reporting and Logging / When to Log  | log_min_duration_statement  | -1
 Reporting and Logging / When to Log  | log_min_error_statement     | error
 Reporting and Logging / When to Log  | log_min_messages            | warning
 Reporting and Logging / When to Log  | log_transaction_sample_rate | 0
 Reporting and Logging / Where to Log | log_destination             | stderr
 Reporting and Logging / Where to Log | log_filename                | postgresql-%Y-%m-%d_%H%M%S.log
 Reporting and Logging / Where to Log | logging_collector           | off
 Reporting and Logging / Where to Log | log_truncate_on_rotation    | off
 Reporting and Logging / Where to Log | log_rotation_size           | 10240
 Reporting and Logging / Where to Log | log_file_mode               | 0600
 Reporting and Logging / Where to Log | log_rotation_age            | 1440
 Reporting and Logging / Where to Log | log_directory               | log
 Statistics / Monitoring              | log_statement_stats         | off
 Statistics / Monitoring              | log_planner_stats           | off
 Statistics / Monitoring              | log_executor_stats          | off
 Statistics / Monitoring              | log_parser_stats            | off
(29 rows)
[local:/var/run/test]:5000 test@testdb=#

可以看到，日志直接輸出在控制臺上，而log_打頭的參數有29個，下面從where、when、what這幾個維度來解析這些參數，本節是第一部分，介紹where。

log：where

與”where”這個維度有所相關的日志參數包括log_directory、log_destination、log_filename、log_file_mode。
log_directory
默認值為log，是一個相對路徑，實際的絕對路徑是$PGDATA/log，當然也可以使用/打頭指定日志存儲的絕對路徑。

log_destination
可選項包括stderr, csvlog, syslog, eventlog
stderr
默認值STDERR,標準輸出（standard error）,上面我們執行查詢配置參數的語句實際結果輸出前的那一串其實是輸出到STDERR的結果.在logging_collector=off的情況下，使用此選項會把日志信息輸出到terminal中。

syslog
操作系統日志daemon.與syslog相關的參數還有以下幾個:
syslog_facility = ‘LOCAL0’,表示 “category of source”
syslog_ident = ‘postgres’,日志的識別符
syslog_sequence_numbers = on,是否開啟序列編號
syslog_split_messages = on,是否拆分消息

下面是修改為syslog后，執行select 1/0后的日志輸出。

[root@localhost log]# uname -a
Linux localhost.localdomain 3.10.0-693.el7.x86_64 #1 SMP Tue Aug 22 21:09:27 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux
[root@localhost log]# pwd
/var/log
[root@localhost log]# tail -n 20 messages
...
Dec 26 11:28:03 localhost postgres[5035]: [7-1] 2019-12-26 11:28:03.737 CST [5035] ERROR:  division by zero
Dec 26 11:28:03 localhost postgres[5035]: [7-2] 2019-12-26 11:28:03.737 CST [5035] STATEMENT:  select 1/0;
...
[root@localhost log]#

postgres是syslog_ident定義的字符串，可以由用戶自行定義，[5035]中的5035是pid。
比如把該參數修改為12.1：

###修改參數
[test@localhost tmpdb]$ pg_ctl reload
server signaled
[test@localhost tmpdb]$ grep 'syslog_ident' postgresql.conf 
syslog_ident = 'pg12.1'
[test@localhost tmpdb]$ 
###日志輸出
Dec 26 11:34:32 localhost postgres[4800]: [7-1] 2019-12-26 11:34:32.906 CST [4800] LOG:  received SIGHUP, reloading configuration files
Dec 26 11:34:32 localhost pg12.1[4800]: [8-1] 2019-12-26 11:34:32.907 CST [4800] LOG:  parameter "syslog_ident" changed to "pg12.1"
Dec 26 11:34:37 localhost pg12.1[5035]: [8-1] 2019-12-26 11:34:37.163 CST [5035] ERROR:  division by zero
Dec 26 11:34:37 localhost pg12.1[5035]: [8-2] 2019-12-26 11:34:37.163 CST [5035] STATEMENT:  select 1/0;
[root@localhost log]#

其中4800是postmaster進程，5035是backend進程
通過ps -ef查看進程信息

[test@localhost tmpdb]$ ps -ef|grep postgres|grep 4800
test      4800     1  0 11:24 ?        00:00:00 /appdb/pg12/pg12.1/bin/postgres
test      4802  4800  0 11:24 ?        00:00:00 postgres: checkpointer   
test      4803  4800  0 11:24 ?        00:00:00 postgres: background writer   
test      4804  4800  0 11:24 ?        00:00:00 postgres: walwriter   
test      4805  4800  0 11:24 ?        00:00:00 postgres: autovacuum launcher  
test      4806  4800  0 11:24 ?        00:00:00 postgres: stats collector   
test      4807  4800  0 11:24 ?        00:00:00 postgres: logical replication launcher  
test      5035  4800  0 11:27 ?        00:00:00 postgres: test testdb [local] idle
[test@localhost tmpdb]$

修改rsyslog的配置，設定local0日志的輸出并重啟rsyslog服務

[root@localhost ~]# grep 'local0' /etc/rsyslog.conf 
local0.*                                                /var/log/postgres.log
[root@localhost ~]#

修改PG日志輸出為syslog，查看日志輸出

[root@localhost ~]# ls -l /var/log/postg*
-rw------- 1 root root 1656 Dec 26 16:29 /var/log/postgres.log
[root@localhost ~]# cat /var/log/postgres.log
Dec 26 16:29:04 localhost pg12.1[23642]: [1-1] 2019-12-26 16:29:04.668 CST [23642] LOG:  starting PostgreSQL 12.1 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-16), 64-bit
Dec 26 16:29:04 localhost pg12.1[23642]: [2-1] 2019-12-26 16:29:04.668 CST [23642] LOG:  listening on IPv6 address "::1", port 5000
Dec 26 16:29:04 localhost pg12.1[23642]: [3-1] 2019-12-26 16:29:04.668 CST [23642] LOG:  listening on IPv4 address "127.0.0.1", port 5000
Dec 26 16:29:04 localhost pg12.1[23642]: [4-1] 2019-12-26 16:29:04.669 CST [23642] LOG:  listening on Unix socket "/var/run/test/.s.PGSQL.5000"
Dec 26 16:29:04 localhost pg12.1[23642]: [5-1] 2019-12-26 16:29:04.683 CST [23642] LOG:  redirecting log output to logging collector process
Dec 26 16:29:04 localhost pg12.1[23642]: [5-2] 2019-12-26 16:29:04.683 CST [23642] HINT:  Future log output will appear in directory "log".
Dec 26 16:29:04 localhost pg12.1[23642]: [6-1] 2019-12-26 16:29:04.684 CST [23642] LOG:  ending log output to stderr
Dec 26 16:29:04 localhost pg12.1[23642]: [6-2] 2019-12-26 16:29:04.684 CST [23642] HINT:  Future log output will go to log destination "syslog".
Dec 26 16:29:04 localhost pg12.1[23644]: [7-1] 2019-12-26 16:29:04.688 CST [23644] LOG:  database system was shut down at 2019-12-26 16:29:04 CST
Dec 26 16:29:04 localhost pg12.1[23642]: [7-1] 2019-12-26 16:29:04.693 CST [23642] LOG:  database system is ready to accept connections
Dec 26 16:29:38 localhost pg12.1[23683]: [8-1] 2019-12-26 16:29:38.824 CST [23683] ERROR:  division by zero
Dec 26 16:29:38 localhost pg12.1[23683]: [8-2] 2019-12-26 16:29:38.824 CST [23683] STATEMENT:  select 1/0;
[root@localhost ~]#

syslog需要依賴平臺的實現，更深入的內容后續再行介紹。

csvlog
csv格式文件。如使用該選項，則需要設置logging_collector參數為on。
設置該參數為on，log_destination設置為stderr

###logging_collector=off
[root@localhost 5035]# ps -ef|grep 4800
test      4800     1  0 11:24 ?        00:00:00 /appdb/pg12/pg12.1/bin/postgres
test      4802  4800  0 11:24 ?        00:00:00 postgres: checkpointer   
test      4803  4800  0 11:24 ?        00:00:00 postgres: background writer   
test      4804  4800  0 11:24 ?        00:00:00 postgres: walwriter   
test      4805  4800  0 11:24 ?        00:00:00 postgres: autovacuum launcher  
test      4806  4800  0 11:24 ?        00:00:00 postgres: stats collector   
test      4807  4800  0 11:24 ?        00:00:00 postgres: logical replication launcher  
test      5035  4800  0 11:27 ?        00:00:00 postgres: test testdb [local] idle
root     15986  5275  0 14:36 pts/5    00:00:00 grep --color=auto 4800
###logging_collector=on
[root@localhost 5035]# ps -ef|grep 16024
test     16024     1  0 14:37 ?        00:00:00 /appdb/pg12/pg12.1/bin/postgres
test     16025 16024  0 14:37 ?        00:00:00 postgres: logger   
test     16027 16024  0 14:37 ?        00:00:00 postgres: checkpointer   
test     16028 16024  0 14:37 ?        00:00:00 postgres: background writer   
test     16029 16024  0 14:37 ?        00:00:00 postgres: walwriter   
test     16030 16024  0 14:37 ?        00:00:00 postgres: autovacuum launcher  
test     16031 16024  0 14:37 ?        00:00:00 postgres: stats collector   
test     16032 16024  0 14:37 ?        00:00:00 postgres: logical replication launcher  
root     16090  5275  0 14:38 pts/5    00:00:00 grep --color=auto 16024

新出現了一個進程logger（后續會逐步解讀這部分的源碼），日志輸出到$PGDATA/log目錄下

[test@localhost tmpdb]$ cd log
[test@localhost log]$ ls
postgresql-2019-12-26_143730.log
[test@localhost log]$ tail -n 20 postgresql-2019-12-26_143730.log 
2019-12-26 14:37:30.317 CST [16026] LOG:  database system was shut down at 2019-12-26 14:37:30 CST
2019-12-26 14:37:30.354 CST [16024] LOG:  database system is ready to accept connections
2019-12-26 14:41:27.684 CST [16293] ERROR:  division by zero
2019-12-26 14:41:27.684 CST [16293] STATEMENT:  select 1/0;
[test@localhost log]$

接著，我們修改該參數為csvlog，執行select 1/0

[test@localhost tmpdb]$ ls -l ./log/
total 12
-rw------- 1 test test 807 Dec 26 14:45 postgresql-2019-12-26_143730.log
-rw------- 1 test test 676 Dec 26 14:45 postgresql-2019-12-26_144540.csv
-rw------- 1 test test 168 Dec 26 14:45 postgresql-2019-12-26_144540.log
[test@localhost tmpdb]$ tail -n 20 ./log/postgresql-2019-12-26_144540.csv
2019-12-26 14:45:40.781 CST,,,16534,,5e045714.4096,1,,2019-12-26 14:45:40 CST,,0,LOG,00000,"ending log output to stderr",,"Future log output will go to log destination ""csvlog"".",,,,,,,""
2019-12-26 14:45:40.802 CST,,,16536,,5e045714.4098,1,,2019-12-26 14:45:40 CST,,0,LOG,00000,"database system was shut down at 2019-12-26 14:45:40 CST",,,,,,,,,""
2019-12-26 14:45:40.808 CST,,,16534,,5e045714.4096,2,,2019-12-26 14:45:40 CST,,0,LOG,00000,"database system is ready to accept connections",,,,,,,,,""
2019-12-26 14:45:44.258 CST,"test","testdb",16546,"[local]",5e045717.40a2,1,"SELECT",2019-12-26 14:45:43 CST,3/2,0,ERROR,22012,"division by zero",,,,,,"select 1/0;",,,"psql"
[test@localhost tmpdb]$ tail -n 20 ./log/postgresql-2019-12-26_144540.log
2019-12-26 14:45:40.781 CST [16534] LOG:  ending log output to stderr
2019-12-26 14:45:40.781 CST [16534] HINT:  Future log output will go to log destination "csvlog".
[test@localhost tmpdb]$ tail -n 100 ./log/postgresql-2019-12-26_144540.log
2019-12-26 14:45:40.781 CST [16534] LOG:  ending log output to stderr
2019-12-26 14:45:40.781 CST [16534] HINT:  Future log output will go to log destination "csvlog".
[test@localhost tmpdb]$

可以看到，修改為csvlog后，輸出的日志格式為以逗號分隔的csv文件格式，這種格式的文件可以很方便的導入到數據庫（Oracle、PG、MySQL等）中。

eventlog
Windows平臺的事件日志。

logging_collector
參見上述解釋，簡單來說，設置為on，則PG會啟動logger進程，同時日志會發送給pg的syslogger進程，否則不會。

log_filename
默認值為’postgresql-%Y-%m-%d_%H%M%S.log’，其中%Y代表年，%m代表月，%d代表日，%H代表小時，%M代表分鐘，%S代表秒。
可用的匹配符可參見 strftime，如使用默認值，則每隔一秒就會產生一個日志文件（如有的話）。

%a
The abbreviated name of the day of the week according to the current locale. (Calculated from tm_wday.)
%A
The full name of the day of the week according to the current locale. (Calculated from tm_wday.)
%b
The abbreviated month name according to the current locale. (Calculated from tm_mon.)
%B
The full month name according to the current locale. (Calculated from tm_mon.)
%c
The preferred date and time representation for the current locale.
%C
The century number (year/100) as a 2-digit integer. (SU) (Calculated from tm_year.)
%d
The day of the month as a decimal number (range 01 to 31). (Calculated from tm_mday.)
%D
Equivalent to %m/%d/%y. (Yecch—for Americans only. Americans should note that in other countries %d/%m/%y is rather common. This means that in international context this format is ambiguous and should not be used.) (SU)
%e
Like %d, the day of the month as a decimal number, but a leading zero is replaced by a space. (SU) (Calculated from tm_mday.)
%E
Modifier: use alternative format, see below. (SU)
%F
Equivalent to %Y-%m-%d (the ISO 8601 date format). (C99)
%G
The ISO 8601 week-based year (see NOTES) with century as a decimal number. The 4-digit year corresponding to the ISO week number (see %V). This has the same format and value as %Y, except that if the ISO week number belongs to the previous or next year, that year is used instead. (TZ) (Calculated from tm_year, tm_yday, and tm_wday.)
%g
Like %G, but without century, that is, with a 2-digit year (00–99). (TZ) (Calculated from tm_year, tm_yday, and tm_wday.)
%h
Equivalent to %b. (SU)
%H
The hour as a decimal number using a 24-hour clock (range 00 to 23). (Calculated from tm_hour.)
%I
The hour as a decimal number using a 12-hour clock (range 01 to 12). (Calculated from tm_hour.)
%j
The day of the year as a decimal number (range 001 to 366). (Calculated from tm_yday.)
%k
The hour (24-hour clock) as a decimal number (range 0 to 23); single digits are preceded by a blank. (See also %H.) (Calculated from tm_hour.) (TZ)
%l
The hour (12-hour clock) as a decimal number (range 1 to 12); single digits are preceded by a blank. (See also %I.) (Calculated from tm_hour.) (TZ)
%m
The month as a decimal number (range 01 to 12). (Calculated from tm_mon.)
%M
The minute as a decimal number (range 00 to 59). (Calculated from tm_min.)
%n
A newline character. (SU)
%O
Modifier: use alternative format, see below. (SU)
%p
Either “AM” or “PM” according to the given time value, or the corresponding strings for the current locale. Noon is treated as “PM” and midnight as “AM”. (Calculated from tm_hour.)
%P
Like %p but in lowercase: “am” or “pm” or a corresponding string for the current locale. (Calculated from tm_hour.) (GNU)
%r
The time in a.m. or p.m. notation. In the POSIX locale this is equivalent to %I:%M:%S %p. (SU)
%R
The time in 24-hour notation (%H:%M). (SU) For a version including the seconds, see %T below.
%s
The number of seconds since the Epoch, 1970-01-01 00:00:00 +0000 (UTC). (TZ) (Calculated from mktime(tm).)
%S
The second as a decimal number (range 00 to 60). (The range is up to 60 to allow for occasional leap seconds.) (Calculated from tm_sec.)
%t
A tab character. (SU)
%T
The time in 24-hour notation (%H:%M:%S). (SU)
%u
The day of the week as a decimal, range 1 to 7, Monday being 1. See also %w. (Calculated from tm_wday.) (SU)
%U
The week number of the current year as a decimal number, range 00 to 53, starting with the first Sunday as the first day of week 01. See also %V and %W. (Calculated from tm_yday and tm_wday.)
%V
The ISO 8601 week number (see NOTES) of the current year as a decimal number, range 01 to 53, where week 1 is the first week that has at least 4 days in the new year. See also %U and %W. (Calculated from tm_year, tm_yday, and tm_wday.) (SU)
%w
The day of the week as a decimal, range 0 to 6, Sunday being 0. See also %u. (Calculated from tm_wday.)
%W
The week number of the current year as a decimal number, range 00 to 53, starting with the first Monday as the first day of week 01. (Calculated from tm_yday and tm_wday.)
%x
The preferred date representation for the current locale without the time.
%X
The preferred time representation for the current locale without the date.
%y
The year as a decimal number without a century (range 00 to 99). (Calculated from tm_year)
%Y
The year as a decimal number including the century. (Calculated from tm_year)
%z
The +hhmm or -hhmm numeric timezone (that is, the hour and minute offset from UTC). (SU)
%Z
The timezone name or abbreviation.
%+
The date and time in date(1) format. (TZ) (Not supported in glibc2.)
%%
A literal ‘%’ character.

log_file_mode
日志文件的mode，用于控制日志文件的訪問權限，默認為0600

[test@localhost tmpdb]$ cd log
[test@localhost log]$ ll
total 16
-rw------- 1 test test  807 Dec 26 14:45 postgresql-2019-12-26_143730.log
-rw------- 1 test test 1584 Dec 26 14:52 postgresql-2019-12-26_144540.csv
-rw------- 1 test test  168 Dec 26 14:45 postgresql-2019-12-26_144540.log
-rw------- 1 test test  309 Dec 26 14:52 postgresql-2019-12-26_145238.log

設置為0777，重啟，文件為666

[test@localhost tmpdb]$ ll ./log
total 20
-rw------- 1 test test  807 Dec 26 14:45 postgresql-2019-12-26_143730.log
-rw------- 1 test test 1584 Dec 26 14:52 postgresql-2019-12-26_144540.csv
-rw------- 1 test test  168 Dec 26 14:45 postgresql-2019-12-26_144540.log
-rw------- 1 test test  807 Dec 26 14:57 postgresql-2019-12-26_145238.log
-rw-rw-rw- 1 test test  309 Dec 26 14:58 postgresql-2019-12-26_145757.log
[test@localhost tmpdb]$

log_truncate_on_rotation
可選項為on和off，默認為off。
如設置為on，在rotate時，如果文件存在，則截斷文件內容，否則在已存在的文件中追加信息。
關于rotation，PG提供了兩種模式：一種基于大小，另外一種基于時間，對應的參數分別是log_rotation_size和log_rotation_age。

log_rotation_size
單位為KB，默認值為10MB。該參數意思是在大小超過該設定后，生成新的日志文件。

log_rotation_age
單位為分鐘，默認為1d，即1440分鐘。該參數意思是經過這個時長之后，產生新的日志文件。如配置為5分鐘，則5分鐘后產生新的日志文件，文件名稱由log_filename控制。

由于文件名稱由log_filename控制，那么在大小或者時間超過設定，但文件名稱設定為不變（如設定為postgresql-%Y-%m-%d.log），結果會如何呢？

[test@localhost tmpdb]$ grep 'log_filename' postgresql.conf 
log_filename = 'postgresql-%Y-%m-%d.log'  # log file name pattern,
[test@localhost tmpdb]$ grep 'rotation' postgresql.conf 
log_truncate_on_rotation = on   # If on, an existing log file with the
                                # time-driven rotation, not on restarts
                                # or size-driven rotation.  Default is
log_rotation_age = 5            # Automatic rotation of logfiles will
log_rotation_size = 10MB        # Automatic rotation of logfiles will
[test@localhost tmpdb]$ 
[test@localhost tmpdb]$ rm -rf ./log
[test@localhost tmpdb]$ pg_ctl restart
waiting for server to shut down.... done
server stopped
waiting for server to start....2019-12-26 15:10:26.812 CST [18042] LOG:  starting PostgreSQL 12.1 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-16), 64-bit
2019-12-26 15:10:26.812 CST [18042] LOG:  listening on IPv6 address "::1", port 5000
2019-12-26 15:10:26.812 CST [18042] LOG:  listening on IPv4 address "127.0.0.1", port 5000
2019-12-26 15:10:26.814 CST [18042] LOG:  listening on Unix socket "/var/run/test/.s.PGSQL.5000"
2019-12-26 15:10:26.823 CST [18042] LOG:  redirecting log output to logging collector process
2019-12-26 15:10:26.823 CST [18042] HINT:  Future log output will appear in directory "log".
 done
server started
[test@localhost tmpdb]$ ll ./log/
total 4
-rw------- 1 test test 188 Dec 26 15:11 postgresql-2019-12-26.log
[test@localhost tmpdb]$ 
[test@localhost tmpdb]$ ll ./log/
total 4
-rw------- 1 test test 188 Dec 26 15:11 postgresql-2019-12-26.log
[test@localhost tmpdb]$ watch -n1 ls -l ./log
[test@localhost tmpdb]$ ll ./log/
total 4
-rw------- 1 test test 551 Dec 26 15:12 postgresql-2019-12-26.log
[test@localhost tmpdb]$

由于文件已存在，因此不會截斷該文件。同理，如文件已存在，就算時間到了也不會截斷。

在工程實踐上，可通過設置log_filename為按周天或者按月天，這樣可以保證周一到周日每天只有一個文件或者從一號到三十一號號一天只有一個文件，避免出現過多的文件。

參考資料
Understanding postgresql.conf : log
Error Reporting and Logging