PostgreSQL FAQ
note
- 无 Foreign Key Arrays
FOREIGN KEY (EACH ELEMENT OF userids) REFERENCES users,
- https://stackoverflow.com/a/50441059/1870054
- https://commitfest.postgresql.org/17/1252/
- S094
- 不支持 column reordering
- Unsupported Features
- SQL:2016
限制
- db 名字
- 最长 63 byte
- 最多 32767 参数占位 -
?
- 范围为 smallint - https://www.postgresql.org/docs/current/limits.html
unique constraint vs unique index
核心业务语义尽量用 constraint
- unique constraint
unique(tid,entity_id)
->flow_tid_entity_id_key
- 可以延后
- 通过 unique index 实现 - 自动创建
- 附带在 TABLE 上
unique
在 create table 时定义alter table TABLE add unique (tid,rid,cid);
在 create table 之后定义
- 不支持
ADD CONSTRAINT IF NOT EXISTS
- 支持
DROP CONSTRAINT IF EXISTS
- unique index
create unique index on flow(tid,entity_id)
->flow_tid_entity_id_idx
- 可以并发
- 可以 带条件
- 独立 INDEX 概念
- 不能在 create table 时定义 - inline index 概念
- UNIQUE 只是 INDEX 的一个限制
- https://stackoverflow.com/a/6804058/1870054
- add constraint if not exists
- https://stackoverflow.com/questions/23542794
- 都支持 UNIQUE NULLS NOT DISTINCT
- 因为是 INDEX 的能力
TOAST
- TOAST = The Oversized-Attribute Storage Technique
- 一行 2kB - TOAST_TUPLE_THRESHOLD
- 超过先尝试压缩到 2kB - TOAST_TUPLE_TARGET
- 默认 default_toast_compression=pglz, 支持 lz4
- 可按列设置 COMPRESSION
- 压缩不足,则分 chunk 存储到关联的 toast 表
- 超过先尝试压缩到 2kB - TOAST_TUPLE_TARGET
- TOAST 使用 oid 跟踪 - 最多 2^32 = 4 十亿 条记录
- 存储
- PLAIN - 不压缩,无 out-of-line - 超过则异常
- EXTENDED - 压缩+out-of-line
- EXTERNAL - out-of-line
- MAIN - 压缩+尝试尽量不 out-of-line
- 参考
- TOAST storage
- wiki TOAST
-- 默认 EXTENDED
ALTER TABLE users SET STORAGE EXTENDED;
-- 修改 TOAST_TUPLE_TARGET
ALTER TABLE users SET (toast_tuple_target = N);
服务重载
- pg_ctl
su postgres
pg_ctl reload
- sql
psql -U postgres
SELECT pg_reload_conf();
- service
service postgresql restart
维护
- postgres_queries_and_commands.sql - Useful PostgreSQL Queries and Commands
升级
- https://www.xf.is/2019/02/26/convert-postgresql-cluster-to-use-page-checksums/
show data_checksums
可以构造一个包含所有版本的镜像,然后进行升级 - Dockerfile
# https://github.com/tianon/docker-postgres-upgrade
docker run --rm \
-v PGDATAOLD:/var/lib/postgresql/OLD/data \
-v PGDATANEW:/var/lib/postgresql/NEW/data \
tianon/postgres-upgrade:OLD-to-NEW
pg_upgrade --old-datadir /var/lib/pgsql/data/ --new-datadir /var/lib/pgsql/10/data/ \
--old-bindir /usr/bin/ --new-bindir /usr/pgsql-10/bin/
CTID
https://postgresql.verite.pro/blog/2019/04/24/oid-column.html
时区问题
- PG 实际存储的是 UTC 不会存储时区
- 时区信息会用于转换
-- 当前时区
show timezone;
show timezone_abbreviations;
-- 可用时区
SELECT * FROM pg_timezone_names;
-- date 转 tz 应该是正常的
select ('2020-01-02'::date)::TIMESTAMPTZ;
-- session 有效
SET TIME ZONE TO 'UTC';
SET TIMEZONE TO 'UTC';
-- 也可以直接指定 offset
SET timezone=-4;
-- 当前时间
-- UTC 可以使用 localtime
SELECT LOCALTIMESTAMP AT TIME ZONE 'Asia/Shanghai';
-- timestamp 不显示 TZ 信息
SELECT NOW()::TIMESTAMP;
-- current_timestamp 是 timestamptz
-- 可以提取 tz 信息 秒、小时
select current_timestamp,
pg_typeof(current_timestamp),
extract(timezone from current_timestamp::timestamptz),
extract(timezone_h from current_timestamp::timestamptz)
;
-- db 配置
ALTER SYSTEM SET timezone = 'UTC';
-- 修改角色时区
ALTER ROLE my_role SET TIMEZONE = '+1';
NULL 字符 / \0
字符
- PG 不允许字符串包含
\0
- 传入之前替换或用 bytea 存储
- JSON 也不允许包含空字符
unsupported Unicode escape sequence
一般是因为 \u0000
, 替换掉即可, pg 的字符串不支持 \u0000
切换用户
set role user;
切换数据库
- 只能重新链接
密码存储
- pgcrypto
- Hashed Passwords with PostgreSQL's pgcrypto
- https://stackoverflow.com/questions/15733196/where-2x-prefix-are-used-in-bcrypt
-- 生成 Hash
select crypt('12345', gen_salt('bf', 8));
-- 判断密码相等
select crypt('12345', password) = password
-- PG 无法处理 2b - 或者插入时修改为 2a
select *
from users
where username = 'admin'
and regexp_replace(password,'^[$]2b','$2a') = crypt('admin', regexp_replace(salt,'^[$]2b','$2a'));
Calculating and Saving Space in PostgreSQL
- Column Tetris
- HN
- 每种数据类型都有特定的对齐要求 - pg-type
- pg_controldata
- 可以获取数据的族群信息
- Maximum data alignment 显示数据的对齐要求
- Database Object Size Functions
pg_controldata data/
-- 列宽
select pg_column_size('int');
pg_control version number: 1002
Catalog version number: 201707211
Database system identifier: 6502788473953883273
Database cluster state: in production
pg_control last modified: 五 3/ 9 12:04:15 2018
Latest checkpoint location: 0/32F8A88
Prior checkpoint location: 0/32DA0D0
Latest checkpoint's REDO location: 0/32F8A50
Latest checkpoint's REDO WAL file: 000000010000000000000003
Latest checkpoint's TimeLineID: 1
Latest checkpoint's PrevTimeLineID: 1
Latest checkpoint's full_page_writes: on
Latest checkpoint's NextXID: 0:730
Latest checkpoint's NextOID: 25609
Latest checkpoint's NextMultiXactId: 1
Latest checkpoint's NextMultiOffset: 0
Latest checkpoint's oldestXID: 548
Latest checkpoint's oldestXID's DB: 1
Latest checkpoint's oldestActiveXID: 730
Latest checkpoint's oldestMultiXid: 1
Latest checkpoint's oldestMulti's DB: 1
Latest checkpoint's oldestCommitTsXid:0
Latest checkpoint's newestCommitTsXid:0
Time of latest checkpoint: 五 3/ 9 12:04:12 2018
Fake LSN counter for unlogged rels: 0/1
Minimum recovery ending location: 0/0
Min recovery ending loc's timeline: 0
Backup start location: 0/0
Backup end location: 0/0
End-of-backup record required: no
wal_level setting: replica
wal_log_hints setting: off
max_connections setting: 100
max_worker_processes setting: 8
max_prepared_xacts setting: 200
max_locks_per_xact setting: 64
track_commit_timestamp setting: off
Maximum data alignment: 8
Database block size: 8192
Blocks per segment of large relation: 131072
WAL block size: 8192
Bytes per WAL segment: 16777216
Maximum length of identifiers: 64
Maximum columns in an index: 32
Maximum size of a TOAST chunk: 1996
Size of a large-object chunk: 2048
Date/time type storage: 64-bit integers
Float4 argument passing: by value
Float8 argument passing: by value
Data page checksum version: 1
Mock authentication nonce: 32f8310a0cf344f7c1432dd733d3cf6065b748697485724af31fbaf7605f50bc
psql 开启时间记录
\timing
# 直接命令行使用
psql -c '\timing' -c 'select 1'
数组索引
GIN 索引是反向索引(inverted indexes), 适用于包含多个值的情况.
- 支持的操作符
<@
@>
=
&&
数组外键
- 不支持
ERROR: cannot alter type of a column used by a view or rule
必须要先 drop view 再操作, 目前没有比较好的操作方式, 但操作都可以在一个事务中完成
有些修改可以通过直接修改 pg_attribute 来达到目的, 但是非常不建议.
时间戳上的毫秒处理
目前没有比较好的处理方式
-- 将一个毫秒的 ts 转为 timestamp 类型
ALTER TABLE my_info
ALTER COLUMN tstmp TYPE TIMESTAMP USING to_timestamp(tstmp / 1000) + ((tstmp % 1000) || ' milliseconds') :: INTERVAL;
查询语句的最大大小
#define MaxAllocSize ((Size) 0x3fffffff) /* 1 gigabyte - 1 */
IN vs any
- https://stackoverflow.com/a/28995514/1870054
- IN
- Bitmap 扫描
- 数据量大时, 执行时间更慢计划时间更久
- ANY
- 会使用临时表做 JOIN
- 数据量大时, 执行时间更久计划时间更快
CREATE TABLE test (
id BIGINT GENERATED BY DEFAULT AS IDENTITY PRIMARY KEY,
val TEXT
);
-- 插入测试数据
DO
$$
BEGIN
FOR i IN 1..100000 LOOP
INSERT INTO test (val) VALUES ('val#' || i);
END LOOP;
END
$$;
EXPLAIN SELECT *
FROM test
WHERE id IN (1, 2, 3);
EXPLAIN SELECT *
FROM test
WHERE id = ANY (VALUES (1), (2), (3));
# IN
Bitmap Heap Scan on test (cost=12.86..19.97 rows=3 width=40)
Recheck Cond: (id = ANY ('{1,2,3}'::bigint[]))
-> Bitmap Index Scan on test_pkey (cost=0.00..12.86 rows=3 width=0)
Index Cond: (id = ANY ('{1,2,3}'::bigint[]))
# ANY
Nested Loop (cost=0.32..25.00 rows=3 width=40)
-> HashAggregate (cost=0.05..0.08 rows=3 width=4)
Group Key: ""*VALUES*"".column1
-> Values Scan on ""*VALUES*"" (cost=0.00..0.04 rows=3 width=4)
-> Index Scan using test_pkey on test (cost=0.28..8.29 rows=1 width=40)
Index Cond: (id = ""*VALUES*"".column1)
-- 测试 IN
DO
$$
DECLARE
x TEXT = '';
r REFCURSOR;
rec RECORD;
BEGIN
x = '0';
FOR i IN 1..1000 LOOP
x = x || ',' || i;
END LOOP;
OPEN r FOR EXECUTE 'EXPLAIN ANALYSE SELECT *
FROM test
WHERE id IN (' || x || ')';
FOR i IN 1..6 LOOP
FETCH r INTO rec;
RAISE NOTICE 'ROW %', rec;
END LOOP;
END
$$;
-- 测试 ANY
DO
$$
DECLARE
x TEXT;
r REFCURSOR;
rec RECORD;
BEGIN
x = '(0)';
FOR i IN 1..1000 LOOP
x = x || ',(' || i || ')';
END LOOP;
OPEN r FOR EXECUTE 'EXPLAIN ANALYSE SELECT *
FROM test
WHERE id = any (VALUES ' || x || ')';
FOR i IN 1..10 LOOP
FETCH r INTO rec;
RAISE NOTICE 'ROW %', rec;
END LOOP;
END
$$;
重置 schema
DROP SCHEMA public CASCADE;
CREATE SCHEMA public;
-- 9.3+ 可能需要
GRANT ALL ON SCHEMA public TO postgres;
GRANT ALL ON SCHEMA public TO public;
psql 使用连接字符串
psql -Atx "host=localhost port=5432 dbname=taop user=taop" -c 'select current_date'
# -d database
psql -d "host=localhost port=5432 dbname=taop user=taop"
control reached end of trigger procedure without RETURN
使用 trigger 时出现,添加 return null
create or replace function sync_events_trigger() returns trigger
as
$$
BEGIN
-- ...
return null; -- this is important
END;
$$ language plpgsql volatile;
query has no destination for result data
在函数中 select 需要指定输出
- 可以将返回类型指定为 table
- 可以
select into
- 可以 返回
return(select * from tab)
- 可以使用 perform/execute 如果不想要结果
CREATE OR REPLACE FUNCTION tst_dates_func()
RETURNS TABLE( v int) as
$$
select 1;
$$ LANGUAGE sql;
RETURN QUERY (select 1);
RETURN(SELECT dblink_disconnect());
SELECT 1 into result;
RETURN result;
PERFORM select 1;
could not resize shared memory segment "/PostgreSQL.153520683" to 1073812480 bytes: No space left on device
增加 docker 的 shm_size
set-returning functions are not allowed in UPDATE
- 替换 unnest(array_value) ->
array_value[1]
- 替换
(regexp_matches(album, '^6,(?:(.+),)?tv\d+'))[1]
->substring (album FROM '^6,(?:(.+),)?tv\d+')
Tuning
- https://news.ycombinator.com/item?id=28489340
- https://postgresqlco.nf/
- Why Uber Engineering Switched from Postgres to MySQL - 2016
- MySQL used as key-value storage layer of homegrown sharded non-relational database
- https://eng.uber.com/postgres-to-mysql-migration/
- Postgres better for read heavy
- MySQL for write heavy
- Pain Point https://pgdash.io/blog/postgresql-six-not-so-easy-pieces.html
- https://sql-info.de/postgresql/postgres-gotchas.html
Sale Point
- transactional ddl
- JSONB,JSON
- Extension
- Native Array
- window function
- custom type
- MySQL 8
- atomic ddl - 弱于 transactional ddl
- window function
- arbitrary check
- JSON
- https://sql-info.de/mysql/gotchas.html
- https://fromdual.com/mysql-limitations
ON CONFLICT DO UPDATE command cannot affect row a second time
- 插入前进行 dedup
- CTE 先 select distinct
生成列
- 生成列不能被手动指定
- stored - INSERT 或 UPDATE 时写入
- virtual - 实时计算 - 类似 view
generated vs default column
- default
- 可指定
- 不指定使用默认 - 计算 1 次
- 可关联其他列
- generated
- 每次行变化都会生成
remaining connection slots are reserved for non-replication superuser connection
排查连接数问题
Note
- 修改 max_connections 必须重启 pg
- 建议应用指定 application_name
-- 查看当前的最大连接数
show max_connections;
-- 查看连接数主要谁占用
-- idle 多还是 active 多
select state, usename, application_name, datname, count(*)
from pg_stat_activity
group by state, usename, application_name, datname
order by 1, 2, 3;
-- 排查 active 连接
select *
from pg_stat_activity
where state = 'active';
-- 按需 kill 连接
CURRENT_TIMESTAMP vs NOW
- 没有区别
- CURRENT_TIMESTAMP=事物开始时间=transaction_timestamp()=now()
- 输入时的
now
不等于now()
- 其他时间
- statement_timestamp() - STABLE - 语句时间
- clock_timestamp() - VOLATILE - 真正当前时间
ON CONFLICT DO UPDATE command cannot affect row a second time (SQLSTATE 21000)
- insert on conflict 时有重复数据
- 插入之前先去重
column "datlastsysoid" does not exist
- PostgreSQL 15 移除了这个列
- https://github.com/pgmodeler/pgmodeler/issues/1689
- navicat 16 也有问题
select distinct datlastsysoid from pg_database;
查找所有外键
SELECT conrelid::regclass AS table_name,
conname AS foreign_key,
pg_get_constraintdef(oid)
FROM pg_constraint
WHERE contype = 'f'
AND connamespace = 'public'::regnamespace
ORDER BY conrelid::regclass::text, contype DESC;
-- psql
\d table_name
ARRAY/数组 外键
cannot use column reference in DEFAULT expression
- 不能在 default 中使用其他列
- 通过 before insert 触发器实现
lateral vs subquery
- lateral
- PostgreSQL 9.3+
- correlated subquery
- 不产生新的行
- 每行只执行 1 次
LEFT JOIN LATERAL (select 1) t ON TRUE
LATERAL (select 1) t
- 查询可以引用外部列
- subquery
- 只能返回一个值
- 只执行 1 次
union vs union all
- union
- 合并去重
- union all
- 不做处理
外键需要建立索引么?
- pk 和 unique 会自动建立索引
- fk 不会自动建立索引
- 建议建立索引
enum vs check vs fk
- enum
- 不能删除
- 顺序不可变
- 不可以模式匹配
- 实际存储 oridinal
- 默认最长 63byte
- 只有 value
- 存储在 pg_enum
- https://www.postgresql.org/docs/current/datatype-enum.html
- check
- 所见即所得
- fk
- 可以包含 label - 方便生成
- 可以动态变化
- 推荐 FK Value
- 不需要 alter table
-- ENUM
CREATE TYPE valid_colors AS ENUM ('red', 'green', 'blue');
CREATE TABLE t (
color VALID_COLORS
);
-- CHECK
CREATE TABLE t (
colors TEXT CHECK (colors IN ('red', 'green', 'blue'))
);
-- FK ID
CREATE TABLE valid_colors (
id SERIAL PRIMARY KEY NOT NULL,
color TEXT
);
INSERT INTO valid_colors (color) VALUES
('red'),
('green'),
('blue');
CREATE TABLE t (
color_id INTEGER REFERENCES valid_colors (id)
);
-- FK value
CREATE TABLE valid_colors (
value text PRIMARY KEY NOT NULL,
label text
);
INSERT INTO valid_colors (value) VALUES
('red'),
('green'),
('blue');
CREATE TABLE t (
color text REFERENCES valid_colors (value)
);
预加载 table 到内存
ERROR: could not open file "base/5/2704": No such file or directory
数据库少了数据文件
A field with precision 5, scale 4 must round to an absolute value less than 10^1
the database system is in recovery mode the database system is in recovery mode
- 使用 pg_isready 判断是否可用
COLLATE
create index on t (name COLLATE "C");
-- 支持使用 Index
explain select * from t where name like 'W%';
-- 不会用 Index
explain select * from t where name = 'Wener';
-- 会用 Index
explain select * from t where name = 'Wener' collate "C";
-- 修改默认 COLLATE
alter table t alter column name set data type text collate "C";
-- 会用 Index
explain select * from t where name = 'Wener';
- COLLATE "C" 支持前缀过滤索引
- LC_COLLATE
ERROR: could not resize shared memory segment "/PostgreSQL.2692148336" to 1073812480 bytes: No space left on device
- vacuum 时发生
sysctl kernel.shmmax
ls -lash /dev/shm
# PostgreSQL.1489521326
K8S 修改
spec:
containers:
- name: postgres
image: postgres
volumeMounts:
- mountPath: /dev/shm
name: dshm
ports:
- containerPort: 5432
volumes:
- name: dshm
emptyDir:
medium: Memory
Docker
docker run --cap-add=SYS_ADMIN --shm-size 2G postgres
K8S start & shutdown
preStop:
exec:
command: ['/bin/sh', '-c', 'pg_ctl -D /var/lib/postgresql/data/pgdata -w -t 60 stop -m fast']
#command: ['/usr/local/bin/pg_ctl stop -D /var/lib/postgresql/data -w -t 60 -m fast']
LC_MONETERY
caution
- collate 不能创建 db 后修改,可以针对 column 或 table 修改。
show lc_collate;
show lc_monetary;
select *
from pg_settings
where name like 'lc_%';
set lc_monetary to "en_IE.utf8";
select 10::money;
- pg_collation
- https://www.postgresql.org/docs/current/collation.html
- LC_COLLATE
- 默认 en_US.utf8
- 推荐 C
- LC_CTYPE
- LC_MONETERY
name | default | aliyun |
---|---|---|
lc_collate | en_US.utf8 | C |
lc_ctype | en_US.utf8 | en_US.utf8 |
lc_messages | en_US.utf8 | "" |
lc_monetary | en_US.utf8 | C |
lc_numeric | en_US.utf8 | C |
lc_time | en_US.utf8 | C |
- initdb -
--encoding=UTF-8 --lc-collate=C --lc-ctype=C
POSTGRES_INITDB_ARGS: '--encoding=UTF-8 --lc-collate=C --lc-ctype=C'
there is no unique constraint matching given keys for referenced table
外键必须 unique
permission denied for table
- 外键需要 REFERENCES
- https://www.postgresql.org/docs/current/ddl-priv.html
source for a multiple-column UPDATE item must be a sub-SELECT or ROW() expression
不能单个字段
on conflict(value) do update set (label)= (excluded.label);
修改为
on conflict(value) do update set (label,extensions)= (excluded.label,excluded.extensions);
for update nowait - FOR UPDATE cannot be applied to the nullable side of an outer join
- mikroorm LockMode.PESSIMISTIC_WRITE_OR_FAIL
permission denied to set session authorization
-- 获取当前的 superuser
SELECT usename
FROM pg_catalog.pg_user
WHERE usesuper = true;
- ⚠️注意 supabase 的 pg superuser 是 supabase_admin
- supabase_admin 密码和 postgres 相同,登录后
ALTER USER postgres WITH SUPERUSER
- supabase_admin 密码和 postgres 相同,登录后
Windows minimal
- 下载 zip
- 解压
- 得到 pgsql 目录
- 17.0
- zip 250MB
- 解压后 997MB
- pgAdmin 650MB
- bin 175MB
- lib 95MB
- doc 31MB
- share 24MB
- 移除后
- 112MB
- 17.0
rm -rf pgAdmin doc include StackBuilder
rm lib/*.pdb bin/*.pdb
cd bin
rm clusterdb.exe createdb.exe createuser.exe dropdb.exe dropuser.exe ecpg.exe oid2name.exe pg_amcheck.exe pg_archivecleanup.exe pg_basebackup.exe pg_checksums.exe pg_combinebackup.exe pg_config.exe pg_controldata.exe pg_createsubscriber.exe pg_isready.exe pg_receivewal.exe pg_recvlogical.exe pg_resetwal.exe pg_rewind.exe pg_test_fsync.exe pg_test_timing.exe pg_upgrade.exe pg_verifybackup.exe pg_waldump.exe pg_walsummary.exe pgbench.exe reindexdb.exe stackbuilder.exe vacuumdb.exe vacuumlo.exe
cd -
rm -rf share/locale
rm -rf share/doc
ls bin/*.exe | sort
# bin/initdb.exe
# bin/pg_ctl.exe
# bin/pg_dump.exe
# bin/pg_dumpall.exe
# bin/pg_restore.exe
# bin/postgres.exe
# bin/psql.exe
package
cd ..
# gnu tar + zstd ~ 27MB
tar -I 'zstd -20' --exclude='.DS_Store' -cvf pgsql.tar.zst pgsql
# just zip ~ 38MB
zip -r -9 pgsql.zip pgsql -x "*.DS_Store"
init
set PATH=%PATH%;%CD%\pgsql\bin
initdb.exe -D .\data -U postgres --encoding=UTF-8 --lc-collate=C --lc-ctype=C
pg_ctl -D ./data start -l db.log
Shell
export PATH=$PWD/pgsql/bin:$PATH
TZ=Asia/Shanghai initdb -D ./data -U postgres --encoding=UTF-8 --lc-collate=C --lc-ctype=C
pg_ctl -D ./data start -l db.log
export PGDATA=$PWD/data
pg_ctl status
psql -U postgres
pg_ctl stop
echo 'listen_addresses = "*"' >> data/postgresql.conf
echo 'host all all 0.0.0.0/0 md5' >> data/pg_hba.conf
pg_ctl reload