Postgres & Red Hat Cluster Suite
6 likes5,764 views
The document outlines high availability planning with a focus on PostgreSQL and Red Hat Cluster Suite (RHCS) solutions. It discusses clustering options, implementation basics, and hardware selection for creating fault-tolerant database environments. Additionally, it emphasizes the importance of proper configurations and teamwork to ensure successful enterprise database operations.
Postgres & Red Hat Cluster Suite
- 1. © 2014 EnterpriseDB Corporation. All rights reserved. 1 What is Your High Availability Plan? Devrim Gündüz Principal Systems Engineer EnterpriseDB
- 2. © 2014 EnterpriseDB Corporation. All rights reserved. 2 • Introduction to EnterpriseDB • Postgres clustering options • What is Red Hat Cluster Suite? • RHCS solves important High Availability problems • RHCS implementation basics • Tips for selecting hardware • PostgreSQL installation tips for RHCS • Q & A Agenda
- 3. © 2014 EnterpriseDB Corporation. All rights reserved. 3 • PostgreSQL contributor, and advocate for more than 12 years • Responsible for PostgreSQL official YUM repository • Fedora contributor since 2006 • Principal Systems Engineer at EDB since 2011 Your Presenter - Devrim Gündüz
- 4. © 2013 EDB All rights reserved 8.1. 4 Introduction to EDB
- 5. © 2014 EnterpriseDB Corporation. All rights reserved. 5 POSTGRES innovation ENTERPRISE reliability 24/7 support Services & training Enterprise-class features & tools Indemnification Product road-map Control Thousands of developers Fast development cycles Low cost No vendor lock-in Advanced features Enabling commercial adoption of Postgres
- 6. © 2014 EnterpriseDB Corporation. All rights reserved. 6 Postgres Plus Advanced Server Postgres Plus Cloud Database High Availability PerformanceManagement REMOTE DBA 24x7 SUPPORT PROFESSIONAL SERVICES TRAINING EDB Serves All Your Postgres Needs PostgreSQL Security
- 7. © 2014 EnterpriseDB Corporation. All rights reserved. 7 Key Strategic Partners
- 8. © 2014 EnterpriseDB Corporation. All rights reserved. 8 EDB Global 2000 Customers © 2013 EnterpriseDB. Privileged and Confidential. All rights reserved. Accenture Advanced Auto Aetna Airgas AOL AT&T Atos Worldwide BAE Systems Banco do Brasil Barclays Bayer Healthcare Boeing Bouygues Telecom CGI Federal Check Point Software Cisco Systems Citigroup Citrix CME-Commodities CMS Cognizant Technology Community Health CSC Deere & Company Dell Deutsche Börse AG Direccion General del Dongfeng Honda eBay Shinsei Financial Siemens Softbank Sony St Jude Healthcare State Farm Swisscom Syngenta Crop Protection Tata Consultancy Services TD Ameritrade Telefonica Telstra Teradata The GAP Tokio Marine Toyota Union Pacific Railroad Vattenfall AB Vivendi Mobile VMWare Volvo Walt Disney Wipro Xerox Yahoo Yamaha Zeejiand Bafang Zions Bancorp InTouch Technology Intuit JPMorgan Chase KDDI KT Kubota Kyocera LG Electronics Lockheed Martin 8 Malaysia Telecom MasterCard McKesson Michelin Mitsubishi Moodys Mosaic ATM Motorola NEC Eisai EMC Corporation Emerson Electric ENSCO Ericsson Expedia Fujitsu G4S Deposita General Electric (GE) Google HCL Infosystems Hitachi HP HTC Global Services Huwaei Technologies IBM ICICI Lombard Infosys NetApp Systems NTT Nokia Northrop Grumman Nucor ONGEI Panasonic PDVSA Phillips Electronics Phoenix Sage QUALCOMM Raytheon Rite Aid RSA Schneider Electric Science Applications Serco-NA SGS SA
- 9. © 2013 EDB All rights reserved 8.1. 9 Postgres Clustering Options
- 10. © 2014 EnterpriseDB Corporation. All rights reserved. 10 Postgres: Several Clustering Options • No single definition for word “cluster” • Solutions for different problems • Popular clustering alternatives • Where does RHCS stand in here? − Actually “Red Hat High Availability Add-on” Bucardo Multimaster replication PGpool Load balancing, connection pooling PL/Proxy Database partitioning Postgres-XC Write-scalable synchronous multimaster Skytools WAL shipping, queuing, and replication Slony Active-passive (+cascading) replication xDB FM EDB Failover Manager (details on next slide)
- 11. © 2014 EnterpriseDB Corporation. All rights reserved. 11 EDB EDB Failover Manager creates fault tolerant database clusters to minimize downtime when a master database fails by keeping data online in High Availability configurations FAILOVER MANAGER Master Streaming Replica Clients Witness Network / Internet The MASTER is down! I agree! Take control now!
- 12. © 2013 EDB All rights reserved 8.1. 12 Red Hat High Availability Add-on
- 13. © 2014 EnterpriseDB Corporation. All rights reserved. 13 • Open Source Clustering Solution • Supports a wide range of hardware • Developed by Red Hat, with the community • Relatively low cost • No data loss • Requires a Red Hat Enterprise Linux base subscription Red Hat High Availability Add-on
- 14. © 2013 EDB All rights reserved 8.1. 14 Red Hat Cluster Suite Solves Important HA Problems
- 15. © 2014 EnterpriseDB Corporation. All rights reserved. 15 • Minimizes downtime • Automatic failover • Transparent to application • Active/passive clustering • Redundancy − Data − Network − Server & Power However, the Red Hat HA Add-on is not − A replication solution − A multimaster solution Red Hat HA Add-on with Postgres
- 16. © 2013 EDB All rights reserved 8.1. 16 Implementation Basics
- 17. © 2014 EnterpriseDB Corporation. All rights reserved. 17 • Needs team work − DBA, Network, Sysadmin teams • No need to use Global File System (GFS) • Manual editing of cluster.conf − Avoid, or sync them all the time • No auto-starting of Postgres • Postgres' init script compatibility Implementation Basics – Part 1
- 18. © 2014 EnterpriseDB Corporation. All rights reserved. 18 • Quorum disk • Firewall considerations • Benefiting from clusterssh on Linux • Yum groupinstall “High Availability” • Yum groupinstall “High Availability Management” Implementation Basics – Part 2
- 19. © 2014 EnterpriseDB Corporation. All rights reserved. 19 • The “High Availability Management” group − luci, ricci • One web interface for managing cluster and storage • Automated deployment of HA packages • Easy integration with existing cluster • Managing multiple clusters with a single interface • Integration of cluster status and logs Using Conga
- 20. © 2014 EnterpriseDB Corporation. All rights reserved. 20 • Everything is handled by the add-on • Detection of failed node − Split brain scenarios • Failing back − Disabling it to avoid extra downtime • Total downtime: 30-60 seconds Failover
- 21. © 2014 EnterpriseDB Corporation. All rights reserved. 21 Implementation basics Active PostgreSQL Server Passive PostgreSQL Server Private Network Shared Storage
- 22. © 2014 EnterpriseDB Corporation. All rights reserved. 22 Implementation basics Sharded nodes, each backing the others up Shared Storage Private Network Switch Node1 Node2 Node3 Node4
- 23. © 2014 EnterpriseDB Corporation. All rights reserved. 23 Implementation basics Data files and archived xlogs of db1 and db2 (/pg01, /pg02, /pga01, /pga02) Data files and resync'd xlogs of db1-RO and db2-RO (/pg03, /pg04, /pga03, /pga04) Passive: DB1 Active: DB2 Active: DB1 Passive: DB2 Streaming replication Streaming replication Shared Storage Shared Storage Active HS: DB1 Passive HS: DB2 Passive HS: DB1 Active HS: DB2
- 24. © 2013 EDB All rights reserved 8.1. 24 Choosing the Right Hardware
- 25. © 2014 EnterpriseDB Corporation. All rights reserved. 25 • This is probably the most important part • You must consult the Red Hat Hardware Compatibility Guide https://access.redhat.com/kb/docs/DOC-30004 • You have to consider firmware • You must have a fencing device − What is fencing? − Importance of fencing − Well-known fencing devices − HP ILO − DELL DRAC − IBM BladeCenter − Cisco UCS − VMware Choosing the Right Hardware
- 26. © 2014 EnterpriseDB Corporation. All rights reserved. 26 • Multicast support • Internal network setup • Enable ICMP echo echo "net.ipv4.icmp_echo_ignore_broadcasts = 0" >> /etc/sysctl.conf • Fiber Channel Switch for storage • SAN − RAID arrays − What about NFS? DAS? NAS? • At least 2 NICs − Why is 4 preferred? Choosing the Right Hardware
- 27. © 2013 EDB All rights reserved 8.1. 27 Postgres Installation tips
- 28. © 2014 EnterpriseDB Corporation. All rights reserved. 28 • Postgres version • Changing Postgres parameters: − listen_addresses − unix_domain_socket − external_pid_file • What about using the passive server as a replica? Postgres Installation Tips
- 29. © 2014 EnterpriseDB Corporation. All rights reserved. 29 • High Availability is important for most enterprise databases • Red Hat Cluster Suite solves important HA problems • Best practices for RHCS implementation − Assemble the right team − Choose appropriate hardware − Fencing is a critical consideration − Configure your network − Configure Postgres for your RHCS infrastructure • Walk through each step carefully to save time and aggravation • Get Help! EDB can help you implement an enterprise-grade RHCS solution Summary
- 30. © 2013 EDB All rights reserved 8.1. 30 Q&A For further information, contact [email protected]