Ali cloud RDS PG, after a few months and cloud-based ECS disk rebuild, features a large revision. 10,11 ECS + cloud disc multi-zone high-availability version is available after the launch of the revision.
Storage upheaval, 25000 iops raised to 1 million iops, 2TB upgrade to 32TB.
1, support ESSD, up to one million IOPS, 32TB capacity while unit prices unchanged.
2, while the storage space and computing specifications complete unbundling, how you want to match how collocation, 4 core specifications can also buy 32TB capacity.
safer
1, support for cloud storage disk encryption. Improve data storage security.
2, support SSL link, improved data transmission security.
3, support more risk control points, risky operations need SMS confirmation. Improve operational safety.
Reliability, Availability, users have the final say
It supports three modes of data protection (maximum protection, the highest available, highest performance). User traffic can be more tradeoff between data reliability, availability, select the corresponding mode. Each mode can be kept in a self service handover.
More intimate version management
Support arbitrarily small version management.
Detailed description
Available main region can be defined
The main support area available options, available services and databases on the same area, to ensure the lowest latency network.
After the switchover occurs instance, if the selected main zone is available when later inconsistent, will issue an alarm, the user can simultaneously available back to the primary zone.
2 supports two versions of PG 10 11
Currently only supports 10, 11. The future will be quick to support the new version, such as 12.
3 dual storage, SSD storage solution with specifications and tied ESSD
Support ESSD storage, up to 1 million IOPS.
Normal cloud disk maximum 6TB, ESSD maximum 32TB.
Examples of storage space at the same time and complete specifications unbundling, 4 cores can sell the space of 32T.
4 data storage more secure, cloud-disk encryption, support BYOK
Data support the overall disk storage, encryption KEY can be provided by the user, it can also be generated by the KMS service.
Instructions:
1, check availability zones (Hong Kong), create encryption KEY. As of press time, being only the Hong Kong REGION KMS support encryption.
https://kms.console.aliyun.com/cn-hongkong/key/list
KEY can create aliases such as rds_pg_key1, do not put the encryption RDS PG KEY deleted, otherwise it will lead RDS PG can not decrypt the data can not be used.
2、授权PostgreSQL请求获取访问您云资源的权限
PostgreSQL请求获取访问您云资源的权限
下方是系统创建的可供PostgreSQL使用的角色，授权后，PostgreSQL拥有对您云资源相应的访问权限。
AliyunPostgreSQLInstanceEncryptionRole
描述： PostgreSQL默认使用此角色来访问KMS
权限描述： 用于 PostgreSQL 服务实例加密角色的授权策略
3、确认已授权 AliyunPostgreSQLInstanceEncryptionRole
https://ram.console.aliyun.com/roles
完成以上3步之后，就可以在购买实例时使用云盘加密。
支持只读锁定功能
例如，用户要做割接，迁移，或者要做一些维护任务时。想把数据库置为只读，避免有数据写入。
将实例设置为只读，可以确保不会有数据写入实例。同时还能继续为只读的业务提供服务。
5 支持任意小版本管理
只要是打包了的小版本，都可以升级、可以回退小版本。
例如有一些业务可能会限定小版本，有了版本管理的功能，再也不用担心回不到以前的版本了。
6 支持数据可靠性、可用性级别配置（保护模式配置）
RDS PG内核支持了自动降级和自动升级功能。
用户可选三种数据保护模式：
1、最大保护，主从复制采用全同步模式，确保主从切换时，数据0丢失（RPO=0）
2、最高可用，从库正常的情况下，采用同步复制模式。从库异常时，自动降级为异步模式，保证可用性。
这种模式确保尽可能的数据0丢失，降级后则与异步模式一样，主从切换如果发生在异步模式时，最多丢失60S。主从切换如果发生在同步模式时，数据0丢失。
3、最高性能，主从采用异步模式同步数据，主从切换如果发生在异步模式时，最多丢失60S。
用户可以根据业务对数据库的实际需求选择几种模式。
7 健壮的架构
1、保证SLB的主节点与用户选择的主可用区一致，确保用户到SLB再到数据库主节点的网络链路最优。
2、云盘与实例主可用区一致，确保IO链路的延迟最低。
3、云盘快照备份，保证了备份的高效率。支持增量备份，最大可以支持32TB数据。
4、备份数据最终会上传到OSS集群，OSS集群与存储集群完全独立，确保了备份数据的冗余性，提供备份的可靠性。
8 SSL数据链路与存储加密 ， 安全双管齐下
9 丰富的监控指标
tps
操作行数
连接
膨胀点
慢查询
长事务
2PC
空间使用
主从延迟，SLOT延迟
数据盘IOPS
数据盘吞吐
数据盘 空间、INODE 使用率
CPU
网络
内存
10 告警默认指标更加丰富有效
CPU使用率
inode使用率
iops使用率
平均活跃连接
磁盘使用率
连接使用率
11 备份更高效、省成本
由于使用了存储快照备份，所以
1、更快
2、更省空间（块级增量备份）
3、恢复更快
4、克隆更方便
12 克隆时可以更换可用区、更新磁盘类型（ssd、essd）
如果一开始使用的是SSD，想更换到ESSD，克隆时可以更换。
克隆时，还可以更换可用区、规格等信息。
13 用户可自定义参数更加安全
虽然RDS PG 11的默认参数更加合理，但是为了满足用户定制化的需求，依旧会开放一些参数的设定。
筛选了用户最常见的自定义参数，同时对参数的可设置范围进行了安全限定，大幅度降低用户因为参数设置有误导致的问题。
可以连到数据库获取当前参数信息：
postgres=> select name,setting,unit from pg_settings where source <>'default';
name | setting | unit
--------------------------------------+----------------------------------------------------------------------------+------
application_name | psql |
archive_command | |
archive_mode | on |
archive_timeout | 300 | s
auth_delay.milliseconds | 3000 | ms
auto_explain.log_analyze | off |
auto_explain.log_buffers | off |
auto_explain.log_format | text |
auto_explain.log_min_duration | -1 | ms
auto_explain.log_nested_statements | off |
auto_explain.log_timing | on |
auto_explain.log_triggers | off |
auto_explain.log_verbose | off |
auto_explain.sample_rate | 1 |
autovacuum | on |
autovacuum_analyze_scale_factor | 0.05 |
autovacuum_freeze_max_age | 1200000000 |
autovacuum_max_workers | 5 |
autovacuum_multixact_freeze_max_age | 1400000000 |
autovacuum_vacuum_cost_delay | 0 | ms
autovacuum_vacuum_cost_limit | 10000 |
autovacuum_vacuum_scale_factor | 0.02 |
autovacuum_work_mem | 418816 | kB
bgwriter_delay | 10 | ms
bgwriter_lru_maxpages | 1000 |
bgwriter_lru_multiplier | 10 |
checkpoint_completion_target | 0.4 |
checkpoint_timeout | 1500 | s
client_encoding | UTF8 |
config_file | /u01/pg5432/data/postgresql.conf |
data_checksums | on |
data_directory | /u01/pg5432/data |
deadlock_timeout | 1000 | ms
default_transaction_deferrable | off |
effective_cache_size | 1572864 | 8kB
effective_io_concurrency | 0 |
enable_partitionwise_aggregate | on |
enable_partitionwise_join | on |
extra_float_digits | 0 |
hba_file | /u01/pg5432/data/pg_hba.conf |
hot_standby | on |
hot_standby_feedback | off |
huge_pages | off |
ident_file | /u01/pg5432/data/pg_ident.conf |
idle_in_transaction_session_timeout | 3600000 | ms
jit | off |
jit_provider | llvmjit |
lc_collate | C |
lc_ctype | en_US.utf8 |
listen_addresses | * |
lock_timeout | 0 | ms
log_autovacuum_min_duration | 0 | ms
log_checkpoints | on |
log_connections | off |
log_destination | csvlog |
log_disconnections | off |
log_duration | off |
log_error_verbosity | verbose |
log_filename | postgresql-%Y-%m-%d_%H%M%S.log |
log_lock_waits | on |
log_min_duration_statement | 5000 | ms
log_rotation_age | 0 | min
log_rotation_size | 102400 | kB
log_statement | ddl |
log_temp_files | 131072 | kB
log_timezone | UTC |
log_truncate_on_rotation | on |
logging_collector | on |
maintenance_work_mem | 1048576 | kB
max_connections | 500 |
max_logical_replication_workers | 56 |
max_parallel_maintenance_workers | 2 |
max_parallel_workers | 2 |
max_parallel_workers_per_gather | 2 |
max_prepared_transactions | 500 |
max_replication_slots | 56 |
max_stack_depth | 2048 | kB
max_standby_archive_delay | 300000 | ms
max_standby_streaming_delay | 300000 | ms
max_sync_workers_per_subscription | 2 |
max_wal_senders | 64 |
max_wal_size | 16384 | MB
max_worker_processes | 256 |
min_wal_size | 2048 | MB
old_snapshot_threshold | -1 | min
pg_pathman.enable | on |
pg_pathman.enable_auto_partition | on |
pg_pathman.enable_bounds_cache | on |
pg_pathman.enable_partitionfilter | on |
pg_pathman.enable_partitionrouter | off |
pg_pathman.enable_runtimeappend | on |
pg_pathman.enable_runtimemergeappend | on |
pg_pathman.insert_into_fdw | postgres |
pg_pathman.override_copy | on |
pg_stat_statements.max | 5000 |
pg_stat_statements.save | off |
pg_stat_statements.track | top |
pg_stat_statements.track_utility | off |
port | 5432 |
random_page_cost | 1.1 |
rds_sync_replication_timeout | 0 | ms
server_encoding | UTF8 |
shared_buffers | 524288 | 8kB
shared_preload_libraries | pg_stat_statements,auth_delay,auto_explain,zhparser,timescaledb,pg_pathman |
ssl | off |
ssl_cert_file | server.crt |
ssl_key_file | server.key |
statement_timeout | 0 | ms
superuser_reserved_connections | 100 |
synchronous_commit | off |
synchronous_standby_names | standby1 |
tcp_keepalives_count | 10 |
tcp_keepalives_idle | 45 | s
tcp_keepalives_interval | 10 | s
temp_file_limit | 16777216 | kB
track_activity_query_size | 1024 | B
track_commit_timestamp | off |
track_functions | pl |
track_io_timing | on |
transaction_deferrable | off |
transaction_isolation | read committed |
transaction_read_only | off |
unix_socket_directories | . |
unix_socket_permissions | 0700 |
vacuum_cost_delay | 0 | ms
vacuum_cost_limit | 10000 |
vacuum_defer_cleanup_age | 0 |
vacuum_freeze_table_age | 200000000 |
vacuum_multixact_freeze_table_age | 200000000 |
wal_buffers | 2048 | 8kB
wal_compression | on |
wal_keep_segments | 0 |
wal_level | replica |
wal_receiver_status_interval | 1 | s
wal_receiver_timeout | 30000 | ms
wal_segment_size | 16777216 | B
wal_writer_delay | 10 | ms
work_mem | 4096 | kB
zhparser.dict_in_memory | off |
zhparser.dicts_type | EXTRA |
zhparser.extra_dicts | dict_extra.xdb |
zhparser.multi_duality | off |
zhparser.multi_short | off |
zhparser.multi_zall | off |
zhparser.multi_zmain | off |
zhparser.punctuation_ignore | off |
zhparser.seg_with_duality | off |
(147 rows)
14 增加了更多的风控点
控制台操作和API操作，和用户操作数据库一样，都有可能出现误操作，为了尽可能降低用户误操作带来的损失，我们增加了一些风控点。
在风控点的操作，都必须短信验证才能继续。
15 账号和数据库管理更加的人性化
1、允许在控制台创建任意个超级或普通账号。
2、允许在控制台锁定账号，例如因为员工离职，或者某些情况下要锁定账号而不是删除账号的功能。锁定的账号，账号OWN的对象还在，经过SQL授权可以继续被其他账号访问。
3、允许在控制台删除账号。
4、允许在控制台创建数据库，可以指定数据库的encodding, lc等信息。
16 支持丰富的插件
支持了非常丰富的插件，例如：
1、cube，用于多维向量分析，相似图像、相似向量特征值快速搜索。
2、rum，用于高级的全文检索、数组、文本搜索等。
3、dblink，跨库，跨实例(网络环境互通后支持)的访问。
4、tablefunc，行列变换。
5、timescaledb，时序数据高性能处理。
6、zhparser中文分词。
7、pg_pathman，高性能分区表插件。（PG 12版本开始，内置分区性能也很高。）
8、orafce，兼容ORACLE的包。
9、varbitx，用户画像插件。
10、postgis，最流行的GIS开源插件。
11、ganos，阿里自研，兼容POSTGIS，同时相比POSTGIS做了大量的功能和性能优化，目前免费使用，特别是在MOD领域，相比开源性能提升50倍以上。
postgres=> show rds_supported_extensions;
rds_supported_extensions
------------------------------------
plpgsql +
pg_stat_statements +
btree_gin +
btree_gist +
citext +
cube +
rum +
dblink +
dict_int +
earthdistance +
hstore +
intagg +
intarray +
isn +
ltree +
pgcrypto +
pgrowlocks +
pg_prewarm +
pg_trgm +
postgres_fdw +
sslinfo +
tablefunc +
timescaledb +
unaccent +
fuzzystrmatch +
pgstattuple +
pg_buffercache +
zhparser +
pg_pathman +
plperl +
pltcl +
"uuid-ossp" +
orafce +
pg_concurrency_control +
varbitx +
postgis +
postgis_sfcgal +
postgis_topology +
postgis_tiger_geocoder +
address_standardizer +
address_standardizer_data_us +
ogr_fdw +
ganos_geometry_sfcgal +
ganos_geometry_topology +
ganos_geometry +
ganos_networking +
ganos_pointcloud_geometry +
ganos_pointcloud +
ganos_raster +
ganos_spatialref +
ganos_trajectory +
ganos_tiger_geocoder +
ganos_address_standardizer +
ganos_address_standardizer_data_us
(1 row)
postgres=> show shared_preload_libraries ;
shared_preload_libraries
----------------------------------------------------------------------------
pg_stat_statements,auth_delay,auto_explain,zhparser,timescaledb,pg_pathman
(1 row)
有一些用户抱怨RDS PG的日志不友好，我们会在RDS PG上支持性能洞察，另外近期会推出log_fdw的插件，用户可以使用LOG_FDW将原始的CSV LOG取走。
Just do it
Based on this framework will continue to evolve in the future, support more users need. Rapid iteration of the new version. Support more plug-ins. Support for more network characteristics (for example, by FDW, EXTENSION other products such as docking cloud). Etc., in short, everything in order to better serve users.
Buy links are as follows:
https://postgresql.console.aliyun.com/cn-shanghai/instance
Now supports mainland China, Hong Kong region. Other line region.