摘要
上文介绍了Oracle收集统计信息策略，以及统计信息过期的逻辑，并没有介绍统计信息过期后收集的后台任务。本文会介绍一下，PG统计信息的收集策略，以及统计信息收集的触发阈值；另外PG会回收死堆元，死堆元的回收阈值和统计信息阈值类似。本文并没有介绍统计信息过期后的收集频率和收集方法，只是介绍了统计信息过期过程，或者说统计信息到了应该收集的阈值过程。关于后台任务如何收集过期的统计信息，将和Oracle和MySQL的统计信息收集任务做对比。
结论假设
1、表数据变化信息收集策略，dml ,create as select ,truncate 的表现
- 增删改的变化，会记录到n_tup_ins，n_tup_upd，n_tup_del，n_tup_hot_upd
- create as select 的数据变化，会记录n_tup_ins
- truncate 表，n_live_tup和n_dead_tup记录会更新为0
2、表收集统计信息的判断逻辑，触发统计信息收集的阈值 Autovacuum ANALYZE threshold for a table = autovacuum_analyze_scale_factor * number of tuples + autovacuum_analyze_threshold 3、PG比较特殊，出了收集统计信息，还会触发回收死堆元 表数据的变化量超过阈值之后，autovaccum 就可以进行回收了。 自动回收触发阈值 Autovacuum VACUUM thresold for a table = autovacuum_vacuum_scale_factor * number of tuples + autovacuum_vacuum_threshold insert 触发阈值 Autovacuum VACUUM thresold for a table = autovacuum_vacuum_insert_threshold* number of tuples + autovacuum_vacuum_insert_threshold PG的统计信息记录在文件中，通过内部函数进行读取，这点也是比较特殊。 需要特别注意，触发阈值的计算，应该是dml操作后记录的n_live_tup，本文按照dml之前的行数计算是不准确的。
14beta3版本模拟测试
1、模拟create as select 和truncate 操作。
postgres=# drop table test ;
DROP TABLE
postgres=# create table test as select * from pg_class;
SELECT 398
postgres=# select now();
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:03:36.708725+08
postgres=# select * from pg_stat_user_tables where relname='test';
-[ RECORD 1 ]-------+-------
relid | 16434
schemaname | public
relname | test
seq_scan | 0
seq_tup_read | 0
idx_scan |
idx_tup_fetch |
n_tup_ins | 398
n_tup_upd | 0
n_tup_del | 0
n_tup_hot_upd | 0
n_live_tup | 398
n_dead_tup | 0
n_mod_since_analyze | 398
n_ins_since_vacuum | 398
last_vacuum |
last_autovacuum |
last_analyze |
last_autoanalyze |
vacuum_count | 0
autovacuum_count | 0
analyze_count | 0
autoanalyze_count | 0
--收集万统计信息后，n_mod_since_analyze字段更新为0
postgres=# select now();
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:04:11.683122+08
postgres=# select * from pg_stat_user_tables where relname='test';
-[ RECORD 1 ]-------+------------------------------
relid | 16434
schemaname | public
relname | test
seq_scan | 0
seq_tup_read | 0
idx_scan |
idx_tup_fetch |
n_tup_ins | 398
n_tup_upd | 0
n_tup_del | 0
n_tup_hot_upd | 0
n_live_tup | 398
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 398
last_vacuum |
last_autovacuum |
last_analyze |
last_autoanalyze | 2021-09-25 20:04:12.671321+08
vacuum_count | 0
autovacuum_count | 0
analyze_count | 0
autoanalyze_count | 1
postgres=# select now();
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:08:41.931437+08
---模拟trancate操作
postgres=# truncate table test;
TRUNCATE TABLE
postgres=# select now();
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:08:55.594144+08
--truncate 操作之后，更新n_live_tup和n_dead_tup
postgres=# select * from pg_stat_user_tables where relname='test';
-[ RECORD 1 ]-------+------------------------------
relid | 16434
schemaname | public
relname | test
seq_scan | 0
seq_tup_read | 0
idx_scan |
idx_tup_fetch |
n_tup_ins | 398
n_tup_upd | 0
n_tup_del | 0
n_tup_hot_upd | 0
n_live_tup | 0
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum |
last_analyze |
last_autoanalyze | 2021-09-25 20:04:12.671321+08
vacuum_count | 0
autovacuum_count | 0
analyze_count | 0
autoanalyze_count | 1
2、模拟dml操作
--创建测试表
postgres=# create table test as select * from pg_class;
SELECT 398
postgres=# select now();
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:30:23.112004+08
--查看统计信息
postgres=# select * from pg_stat_user_tables where relname='test';
-[ RECORD 1 ]-------+-------
relid | 16445
schemaname | public
relname | test
seq_scan | 0
seq_tup_read | 0
idx_scan |
idx_tup_fetch |
n_tup_ins | 398
n_tup_upd | 0
n_tup_del | 0
n_tup_hot_upd | 0
n_live_tup | 398
n_dead_tup | 0
n_mod_since_analyze | 398
n_ins_since_vacuum | 398
last_vacuum |
last_autovacuum |
last_analyze |
last_autoanalyze |
vacuum_count | 0
autovacuum_count | 0
analyze_count | 0
autoanalyze_count | 0
postgres=#
--更新398条记录
postgres=# update test set relname =22;
UPDATE 398
postgres=# select * from test ;
-[ RECORD 1 ]-------+-----------------------------------------
oid | 3575
relname | 22
relnamespace | 11
reltype | 0
reloftype | 0
relowner | 10
relam | 403
relfilenode | 3575
reltablespace | 0
relpages | 1
reltuples | 0
relallvisible | 0
reltoastrelid | 0
relhasindex | f
relisshared | f
relpersistence | p
relkind | i
relnatts | 2
relchecks | 0
relhasrules | f
relhastriggers | f
relhassubclass | f
relrowsecurity | f
relforcerowsecurity | f
relispopulated | t
relreplident | n
relispartition | f
relrewrite | 0
relfrozenxid | 0
relminmxid | 0
relacl |
reloptions |
relpartbound |
-[ RECORD 2 ]-------+-----------------------------------------
oid | 13714
relname | 22
relnamespace | 99
--删除210条记录
postgres=# delete from test where oid >4000;
DELETE 210
--可以看到统计信息，插入了398条，更新了398，删除了210，记录数为188
n_mod_since_analyze和n_ins_since_vacuum记录数为0，是因为统计信息收集了。
postgres=# select * from pg_stat_user_tables where relname='test';
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 398
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 188
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 20:33:14.011364+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 2
3、模拟统计信息和autovaccum触发阈值 Autovacuum VACUUM thresold for a table = autovacuum_vacuum_scale_factor * number of tuples + autovacuum_vacuum_threshold Autovacuum VACUUM thresold for a table = autovacuum_vacuum_insert_threshold* number of tuples + >autovacuum_vacuum_insert_threshold Autovacuum ANALYZE threshold for a table = autovacuum_analyze_scale_factor * number of tuples + autovacuum_analyze_threshold autovacuum_analyze_threshold 50 autovacuum_vacuum_threshold 50 autovacuum_vacuum_scale_factor | 0.2 autovacuum_analyze_scale_factor | 0.1 autovacuum_vacuum_insert_scale_factor 0.2 autovacuum_vacuum_insert_threshold | 1000
3.1统计信息收集阈值模拟 Autovacuum ANALYZE threshold for a table = autovacuum_analyze_scale_factor * number of tuples + autovacuum_analyze_threshold
postgres=# select * from pg_stat_user_tables where relname='test';
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 799
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 401
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 401
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 20:45:14.656733+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 3
--触发条件 0.1*401 +50 =90
postgres=# insert into test select * from pg_class limit 91 ; select now();
INSERT 0 91
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:49:51.103724+08
--可以看到n_mod_since_analyze 为91
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 890
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 492
n_dead_tup | 0
n_mod_since_analyze | 91
n_ins_since_vacuum | 492
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 20:45:14.656733+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 3
-[ RECORD 1 ]----------------------
now | 2021-09-25 20:49:53.112268+08
--收集万统计信息后，n_mod_since_analyze更新为0
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 890
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 492
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 492
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 20:50:14.904217+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 4
3.2 模拟autovacuum 触发条件insert 阈值
Autovacuum VACUUM thresold for a table = autovacuum_vacuum_insert_threshold* number of tuples + autovacuum_vacuum_insert_threshold
–0.2*641+100=1208.4 -1042 =166.4 n_ins_since_vacuum 1042值需要大于insert 的触发阈值，所以只需模拟插入166条记录。
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 1440
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 1042
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 1042
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 21:21:16.407244+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 6
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:24:00.909854+08
----模拟插入166
postgres=# insert into test select * from pg_class limit 167 ;select now();
INSERT 0 167
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:24:18.140023+08
--可以看到n_ins_since_vacuum = 1209，达到触发阈值，等待autovacuum回收。
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 1209
n_dead_tup | 0
n_mod_since_analyze | 167
n_ins_since_vacuum | 1209
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 21:21:16.407244+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 6
-[ RECORD 1 ]---------------------
now | 2021-09-25 21:24:18.95219+08
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 1209
n_dead_tup | 0
n_mod_since_analyze | 167
n_ins_since_vacuum | 1209
last_vacuum |
last_autovacuum | 2021-09-25 20:33:14.00916+08
last_analyze |
last_autoanalyze | 2021-09-25 21:21:16.407244+08
vacuum_count | 0
autovacuum_count | 2
analyze_count | 0
autoanalyze_count | 6
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:25:00.312934+08
-- 可以看到autovacuum 执行了，并且同时更新了统计信息，n_ins_since_vacuum变为0
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 1209
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:25:16.546408+08
last_analyze |
last_autoanalyze | 2021-09-25 21:25:16.572833+08
vacuum_count | 0
autovacuum_count | 3
analyze_count | 0
autoanalyze_count | 7
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:25:36.791924+08
3.3 模拟autovacuum 触发条件 ,delete和update 阈值
Autovacuum VACUUM thresold for a table = autovacuum_vacuum_scale_factor * number of tuples + autovacuum_vacuum_threshold 0.2*1209+50=291.8
with t1 as (select ctid from test limit 292) delete from test where ctid in (select ctid from t1); select * from pg_stat_user_tables where relname=’test’; select now();
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 3
seq_tup_read | 1194
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 210
n_tup_hot_upd | 2
n_live_tup | 1209
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:25:16.546408+08
last_analyze |
last_autoanalyze | 2021-09-25 21:25:16.572833+08
vacuum_count | 0
autovacuum_count | 3
analyze_count | 0
autoanalyze_count | 7
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:35:10.474517+08
--删除292条记录
postgres=# with t1 as (select ctid from test limit 292) delete from test where ctid in (select ctid from t1);select now();
DELETE 292
now
-------------------------------
2021-09-25 21:40:58.620179+08
--可以看到n_dead_tup=292
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 7
seq_tup_read | 2715
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 502
n_tup_hot_upd | 2
n_live_tup | 917
n_dead_tup | 292
n_mod_since_analyze | 292
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:25:16.546408+08
last_analyze |
last_autoanalyze | 2021-09-25 21:25:16.572833+08
vacuum_count | 0
autovacuum_count | 3
analyze_count | 0
autoanalyze_count | 7
-[ RECORD 1 ]--------------------
now | 2021-09-25 21:41:02.4531+08
--可以看到触发autovacuum，先更新回收死堆元，然后收集统计信息，n_dead_tup变为0
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 7
seq_tup_read | 2715
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 502
n_tup_hot_upd | 2
n_live_tup | 917
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:41:17.383646+08
last_analyze |
last_autoanalyze | 2021-09-25 21:41:17.395919+08
vacuum_count | 0
autovacuum_count | 4
analyze_count | 0
autoanalyze_count | 8
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:41:33.879876+08
---模拟未触发的场景
Autovacuum VACUUM thresold for a table = autovacuum_vacuum_scale_factor * number of tuples + autovacuum_vacuum_threshold
0.2*917+50=233.4
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 7
seq_tup_read | 2715
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 502
n_tup_hot_upd | 2
n_live_tup | 917
n_dead_tup | 0
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:41:17.383646+08
last_analyze |
last_autoanalyze | 2021-09-25 21:41:17.395919+08
vacuum_count | 0
autovacuum_count | 4
analyze_count | 0
autoanalyze_count | 8
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:48:48.883104+08
postgres=#
postgres=#
postgres=# with t1 as (select ctid from test limit 232) delete from test where ctid in (select ctid from t1);select now();
DELETE 232
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:49:01.016617+08
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 9
seq_tup_read | 3864
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 734
n_tup_hot_upd | 2
n_live_tup | 685
n_dead_tup | 232
n_mod_since_analyze | 232
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:41:17.383646+08
last_analyze |
last_autoanalyze | 2021-09-25 21:41:17.395919+08
vacuum_count | 0
autovacuum_count | 4
analyze_count | 0
autoanalyze_count | 8
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:49:01.835688+08
postgres=# select * from pg_stat_user_tables where relname='test'; select now();
-[ RECORD 1 ]-------+------------------------------
relid | 16445
schemaname | public
relname | test
seq_scan | 9
seq_tup_read | 3864
idx_scan |
idx_tup_fetch |
n_tup_ins | 1607
n_tup_upd | 398
n_tup_del | 734
n_tup_hot_upd | 2
n_live_tup | 685
n_dead_tup | 232
n_mod_since_analyze | 0
n_ins_since_vacuum | 0
last_vacuum |
last_autovacuum | 2021-09-25 21:41:17.383646+08
last_analyze |
last_autoanalyze | 2021-09-25 21:49:17.877156+08
vacuum_count | 0
autovacuum_count | 4
analyze_count | 0
autoanalyze_count | 9
-[ RECORD 1 ]----------------------
now | 2021-09-25 21:49:57.084457+08
4、实例运行的时候统计信息存储在pg_stat_tmp
postgres=# select pg_backend_pid(); -[ RECORD 1 ]–+—— pg_backend_pid | 46073
可以看到回到pg_stat_tmp目录下读取统计信息
strace 46073
[root@master-node ~]# strace -p 46073
strace: Process 46073 attached
epoll_wait(4, [{EPOLLIN, {u32=43557608, u64=43557608}}], 1, -1) = 1
recvfrom(9, "Q\0\0\0<select * from pg_stat_user_"..., 8192, 0, NULL, NULL) = 61
lseek(6, 0, SEEK_END) = 106496
lseek(7, 0, SEEK_END) = 32768
lseek(12, 0, SEEK_END) = 49152
lseek(13, 0, SEEK_END) = 40960
lseek(14, 0, SEEK_END) = 32768
lseek(15, 0, SEEK_END) = 16384
lseek(16, 0, SEEK_END) = 16384
lseek(17, 0, SEEK_END) = 8192
lseek(18, 0, SEEK_END) = 16384
lseek(19, 0, SEEK_END) = 16384
brk(NULL) = 0x2b00000
brk(0x2b40000) = 0x2b40000
lseek(14, 0, SEEK_END) = 32768
lseek(17, 0, SEEK_END) = 8192
open("pg_stat_tmp/global.stat", O_RDONLY) = 49
fstat(49, {st_mode=S_IFREG|0600, st_size=1335, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f855d37c000
read(49, "\242\274\245\1I\345\3264\321o\2\0W\0\0\0\0\0\0\0\1\0\0\0\0\0\0\0003\332\0\0"..., 4096) = 1335
close(49) = 0
munmap(0x7f855d37c000, 4096) = 0
sendto(8, "\1\0\0\0 \0\0\0\361\305M7\321o\2\0\321$F7\321o\2\0D6\0\0\0\0\0\0", 32, 0, NULL, 0) = 32
select(0, NULL, NULL, NULL, {tv_sec=0, tv_usec=10000}) = 0 (Timeout)
open("pg_stat_tmp/global.stat", O_RDONLY) = 49
fstat(49, {st_mode=S_IFREG|0600, st_size=1335, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f855d37c000
read(49, "\242\274\245\1\372\306M7\321o\2\0W\0\0\0\0\0\0\0\1\0\0\0\0\0\0\0003\332\0\0"..., 4096) = 1335
close(49) = 0
munmap(0x7f855d37c000, 4096) = 0
open("pg_stat_tmp/global.stat", O_RDONLY) = 49
fstat(49, {st_mode=S_IFREG|0600, st_size=1335, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f855d37c000
read(49, "\242\274\245\1\372\306M7\321o\2\0W\0\0\0\0\0\0\0\1\0\0\0\0\0\0\0003\332\0\0"..., 4096) = 1335
open("pg_stat_tmp/db_13892.stat", O_RDONLY) = 50
fstat(50, {st_mode=S_IFREG|0600, st_size=40705, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f855d37b000
read(50, "\242\274\245\1T\25\v\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\2"..., 4096) = 4096
brk(NULL) = 0x2b40000
brk(0x2b67000) = 0x2b67000
read(50, "\0\0\0\0\0\236\1\0\0\0\0\0\0\226\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0T\16\v\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\226\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0S\v\0\0\0\0\0\0\226\0\0\0"..., 4096) = 3841
close(50) = 0
munmap(0x7f855d37b000, 4096) = 0
open("pg_stat_tmp/db_0.stat", O_RDONLY) = 50
fstat(50, {st_mode=S_IFREG|0600, st_size=7960, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f855d37b000
read(50, "\242\274\245\1T\37\v\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
read(50, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 3864
close(50) = 0
munmap(0x7f855d37b000, 4096) = 0
close(49) = 0
munmap(0x7f855d37c000, 4096) = 0
brk(NULL) = 0x2b67000
brk(NULL) = 0x2b67000
brk(0x2b00000) = 0x2b00000
brk(NULL) = 0x2b00000
sendto(8, "\31\0\0\08\0\0\0D6\0\0\0\0\0\0\0\0\0\0\0\0\0\0\253\312I\3\0\0\0\0"..., 56, 0, NULL, 0) = 56
sendto(8, "\2\0\0\0\250\3\0\0D6\0\0\10\0\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 936, 0, NULL, 0) = 936
sendto(8, "\2\0\0\0\230\0\0\0D6\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 152, 0, NULL, 0) = 152
sendto(9, "T\0\0\2\313\0\27relid\0\0\0000V\0\1\0\0\0\32\0\4\377\377\377\377\0\0s"..., 936, 0, NULL, 0) = 936
recvfrom(9, 0xd7e400, 8192, 0, NULL, NULL) = -1 EAGAIN (Resource temporarily unavailable)
epoll_wait(4, ^Cstrace: Process 46073 detached
<detached ...>
结论
1、表的增删改产生的数据条目数变化，通过视图pg_stat_user_tables可以查看，记录在文件中。 2、表的统计信息，超过阈值后台进程会重新收集统计信息。 3、比较特殊，表的死堆元达到阈值后，会触发死堆元回收，也会更新统计信息。