patroni-based citrus high availability environment deployment

Patroni-based citus high availability
environment deployment
( 基于Patroni的Citus高可用环境部署 )
PGDay.Seoul & PGConf.Asia 2021
Daniel Lee
2021-11-24 PGDay.Seoul & PGConf.Asia 2021 1

About Me
• Dainel lee ( 李炯彩 )
• Community
• MySQL Korea User Group
• facebook.com/groups/mysqlko
• PostgreSQL Korea
• facebook.com/groups/130019777169719
• Github
• github.com/dbdbdeep-git/pgday-seoul-2021-demo

Agenda
1. Service Discovery ( *Service Mesh )
• CoreOS’s ETCD
• Hashicorp’s Consul
2. Dynamic Configuration Management
• ConfD
• Hashicorp’s Consul-Template
3. High-Availability
• Zalando’s Patroni
4. Distributed PostgreSQL as an extension
• Citusdata’s Citus
5. Demo

1. Service Discovery

What is …
Lock
Service
(chubby)
Coordination
Service
(zookeeper,eureka,etcd)
Service Discovery
(surf,skydns,consul)
Service Mesh
(lstio,linkerd2)

What can you do with Consul?
Service Discovery
for connectivity
Service Registry enables services to
register and discover each other.
Service Segmentation
for security
Secure service-to-service communication
with automatic TLS encryption and
identity-based authorization.
Service Configuration
for runtime configuration
Feature rich Key/Value store to easily
configure services.

What can you do with Consul?
Consul-Kubernetes Deployments
Use Consul service discovery and
service mesh features with Kubernetes.
Secure Service Communication
Secure and observe communication between
your services without modifying their code.
Dynamic Load Balancing
Automate load balancer configuration with
Consul and HAProxy, Nginx, or F5.

3 Marine Admirals
Properties Consul Etcd Zookeeper
User Interface Available X X
RPC Available Available
Health Check HTTP API HTTP API TCP
Key Value
3 Consistency modes
(sync)
Good Consistency
(semi)
Strong Consistency
Token System Available X X
Language Golang Golang Java
Security ACL / HTTPS HTTPS ACL
Spring Cloud O O O
Multi
Datacenter(gossip)
O X X

2. Dynamic Configuration

ConfD
• http://www.confd.io
• Manage local application configuration files using templates and
data from etcd or consul
• A lightweight configuration management tool focused on:
• keeping local configuration files up-to-date using data stored in etcd,
consul, dynamodb, redis, vault, zookeeper, aws ssm parameter store or
env vars and processing template resources.
• reloading applications to pick up new config file changes

ConfD or Consul-template Architecture
ETCD or Consul
Server
Confd or
Consul-template
Service
Application
(HAproxy, PostgreSQL)
Configuration
Template
File
Application
Configuration
File
listen
generate
haproxy.cfg.tmpl haproxy.cfg
reload
check
reload
update

PostgreSQL Client authentication
Consul Server
Consul-template
Service
Postgres Server
Configuration
Template
File
Application
Configuration
File
listen
generate
pg_hba.conf.tmpl pg_hba.conf
reload
check
reload
update
$ consul kv put
service/database/primary/auth/client
“host all haproxy 10.0.1.0/24 reject”

3. Zalando’s Patroni

Zalando Company

’s Patroni
• Patroni is a template for you to create your own customized, high-availability solution using
Python and - for maximum accessibility - a DCS (Distributed Configuration Store)
like ZooKeeper, etcd, Consul or Kubernetes.
• Currently supported PostgreSQL versions: 9.3 to 14.
• Release lastest 2.1.1 ( 09.11.21 )

’s Spilo
• Spilo is a Docker image that provides PostgreSQL and Patroni bundled
together. Patroni is a template for PostgreSQL HA. Multiple Spilos can create
a resilient High Available PostgreSQL cluster. For this, you'll need to start all
participating Spilos with identical etcd addresses and cluster names.
• https://github.com/zalando/spilo

Raft : DC to DC

HA : Patroni vs pg_auto_failover

4. Citusdata’s Citus

Hyperscale (Citus) on Azure Database for PostgreSQL

Citusとは? by Noriyoshi Shinoda
• PostgreSQL でスケールアウト環境を実現
• 複数ノードにまたがったパラレル・クエリーとパーティショニング機能
• スループット拡大を目指す
• PostgreSQL の拡張（EXTENSION）として実装
• PostgreSQL 本体の変更なし
• Citus Data （https://www.citusdata.com/）が開発
• 2019年1月 Microsoft による買収
• Azure Database for PostgreSQL - Hyperscale (Citus) の中核技術となっている
• オープンソース版も提供（https://github.com/citusdata/citus）
• 以下の機能は含まない
• 自動フェイルオーバー
• 自動データ・リバランス
• バックアップ等の運用機能
https://www.slideshare.net/noriyoshishinoda/postgresql-conference-japan-2021-b2-citus-10

Citus 란?
• PostgreSQL로 스케일 아웃 환경 실현
• 다중 노드에 걸친 병렬 쿼리 및 파티셔닝 기능
• 처리량 확대를 목표로
• PostgreSQL 확장(EXTENSION)으로 구현
• PostgreSQL 본체 변경 없음
• Citus Data (https://www.citusdata.com/) 개발
• 2019년 1월 Microsoft에서 인수
• Azure Database for PostgreSQL - Hyperscale (Citus)의 핵심 기술입니다.
• 오픈 소스 버전도 제공 (https://github.com/citusdata/citus)
• 다음 기능은 포함되지 않음
• 자동 장애 조치
• 자동 데이터 재조정
• 백업 등의 운용 기능

What is Citus ?
• Achieve a scale-out environment with PostgreSQL
• Parallel query and partitioning capabilities across multiple nodes
• Aim to increase throughput
• Implemented as an extension of PostgreSQL (EXTENSION)
• No changes to PostgreSQL itself
• Developed by Citus Data (https://www.citusdata.com/)
• Acquired by Microsoft in January 2019
• Azure Database for PostgreSQL-is the core technology of Hyperscale (Citus)
• Open source version is also available (https://github.com/citusdata/citus)
• The following functions are not included
• Automatic failover
• Automatic data rebalancing
• Operational functions such as backup

What is Citus ? by Noriyoshi Shinoda
Instance configuration
• Coordinator Node
• A PostgreSQL instance that accepts connections from clients
• Manage metadata
• Worker Node
• PostgreSQL instance that actually stores the data
• No communication between Worker Nodes
• Install citus EXTENSION on all nodes

What is Citus ? by Noriyoshi Shinoda
Instance configuration
• Processing performed by Coordinator Node
• Accepting connections from clients
• Final sort (ORDER BY)
• Sequence processing (including SERIAL column and GENERATED AS
IDENTITY column)
• Handling objects other than tables and indexes
• Processing executed by Worker Node
• Processing SQL statements requested by Coordinator Node
• Execution of ANALYZE statement
• Execution of VACUUM statement
• Retaining table data

CREATE EXTENSION citus;
• Add nodes:
postgres=> SELECT master_add_node(‘worker-01_ip_or_dns’, 5432);
Postgres=> SELECT master_add_node(’worker-02_ip_or_dns’, 5432);
• Create distributed table:
postgres=> CREATE TABLE events ( orders_id int, ... );
postgres=> SELECT create_distributed_table( ‘orders', ‘orders_id’ );
• Events is now distributed across shards on the nodes

if(kakao) 2021
https://if.kakao.com/session/59

Table structure by Noriyoshi Shinoda
Distributed table
• A table that distributes and stores data
• Suitable for fact tables
• Specify a column as the distribution key
(default determines the distribution destination table by the hash value range)
• The number of divisions can be specified
• citus.shard_count (default value 32)
• Replicas can be created on different Worker Nodes
• citus.shard_replication_factor (default value 1 = no replica)

Table structure by Noriyoshi Shinoda
Distributed table
• Table created in worker Node
• For the table name, ShardID is added to the original table name.
• If a replica is specified, a table with the same name will be created in a different Worker Node.
• Same data is stored in the table with the same name
• Search the citus_shards catalog to see the table names created by the Worker Node
postgres=> SET citus.shard_count = 6 ;
postgres=> SET citus.shard_replication_factor = 2 ;
Coordinator Worker#1 Worker#2 Worker#3
Dist1
dist1_102046 dist1_102046
dist1_102047 dist1_102047
dist1_102048 dist1_102048
dist1_102049 dist1_102049
dist1_102050 dist1_102050
dist1_102051 dist1_102051

Demo | Skip

if(kakao) 2021

Q?!A

patroni-based citrus high availability environment deployment

More Related Content

What's hot (20)

Similar to patroni-based citrus high availability environment deployment (20)

More from hyeongchae lee (12)

Recently uploaded (20)

patroni-based citrus high availability environment deployment