Ingres High Availability Option May 2008
For information contact Product Management at products@ingres.com This presentation contains forward-looking statements that are based on management s expectations, estimates, projections and assumptions. Forward-looking statements are made pursuant to the safe harbor provisions of the Private Securities Litigation Reform Act of 1995, as amended. These statements are not guarantees of future performance or functionality and involve certain risks and uncertainties, which are difficult to predict. Therefore, actual future functionality, features, results and trends may differ materially from what is forecast in forward-looking statements due to a variety of factors.
Abstract This session examines the Ingres High Availability Option for cluster failover. This configuration allows automatic relocation of data services for a failed primary node to a designated secondary node. With the exception of a brief interruption for client reconnection, the physical server on which Ingres runs is transparent to the end user, and application continuity is maintained. This session examines this facility in detail, and describes this option in the context of other strategies for robust information handling, such as data backup and recovery and data replication.
A longer presentation title How to Succeed in Business by Not Really Failing... the Ingres High Availability Option
Agenda Cluster Overview and Terminology High Availability Option Ingres Cluster Solution for OpenVMS Ongoing Development... Other complementary Ingres features
Terminology Computer network any group of computers that are connected, however loosely Computer cluster a group of coupled computers that work together closely so that in many respects they can be viewed as though they are a single computer [Wikipedia]
Cluster Terminology High-availability (HA) clusters Primarily for improved availability of services ( two Redundant nodes (usually Does not improve the scalability of the server Load-balancing clusters Spread workload over multiple peer nodes Grid computing ( network Loosely clustered (more like a generic Best for work that can be broken into independent chunks
Terms we throw around a lot HA clusters Ingres High Availability Option Active-Passive Cluster Failover Cluster Load-balancing clusters Ingres Cluster Solution Active-Active Cluster Grid computing ( workload N/A (because of the nature of an RDBMS Ingres Grid Option... NO!... it is a cluster option
Ingres High Availability Option Integrates with OS vendor HA failover services Provides co-located redundant server with automated system restart on backup server Optionally restart first on main node a n ( platform d (failing that) on backup node (details vary by Available platforms Red Hat Cluster Suite Windows Server Sun Cluster
Typical HA Cluster For a clustered Ingres installation the database, checkpoint, journal, dump, work, and transaction log files must be on a cluster drive.
Where to get info re Ingres HA Option Ingres Installation Guide Platform-specific README files http://www.ingres.com/ http://servicedesk.ingres.com/ Knowledge Base http://community.ingres.com/ comp.databases.ingres IRC irc.freenode.net #ingres
Where to get info re vendor clusters Red Hat Linux Cluster Windows Server Cluster "Sun Cluster Data Services Planning and Administration Guide"
Red Hat Linux Cluster Ingres is a service that is added to the cluster Service can be assigned an IP address, which is failed over to the backup server for transparent client access Client will see failure of transactions that are in-flight at the ti m ( implementations e of failover (true for all failover Scripts (e.g. ha_ingres) are examples to be tailored to a site's needs and individual installation values
Windows Cluster Virtual server Unique network name IP address Ingres is installed on this virtual server Clients use the virtual server name to connect to Ingres ( scripting Cluster Administrator and cluster.exe (for Installation of the HA option is an intermediary step during Ingres installation Must st a (? Why ) rt and stop Ingres through the cluster service!
Windows Installation Wizard
Sun Cluster Ditto most of the points regarding Red Hat and Windows Minor differences (e.g. initial installation is st a ( follows ndard but is on clustered disks; cluster setup IngresSCAgent available from Ingres web site Logical hostname allows a floating IP address Start and stop Ingres using cl u ( ingstop ster service commands (do not use ingstart and
Ingres Cluster Solution An active-active cluster solution Currently only available for VMS VMS provides all active-active required components as an integrated set of services Cluster services Shared disk Distributed Lock Manager Test bed for Ingres cluster for any future platforms
Ongoing development... Development work was done for active-active clustering on Linux but was not released as a product Difficulty of providing integrated solution with separate off-the-shelf components OpenDLM issues Scaling Did provide for backup servers but HA is simpler and thus more reliable if failover is the major goal ( MVCC ) Multiversion concurrency control New development in the lab this spring May help solve the scaling problem
Other complementary Ingres features Classic backup and recovery procedures Replication Ingres Replicator Third-party solutions Because... HA Option inherent limitations OS cluster services Co-location of equipment Power Connectivity Shared disk Disasters come in many forms!
Questions?