Re: Server consolidation and minimizing unplanned downtime

From: <DEEDSD_at_nationwide.com>
Date: Thu, 14 Jan 2010 14:50:17 -0500
Message-ID: <OFC51F06C0.1FAF1889-ON852576AB.006BCCF3-852576AB.006CF8E3_at_lnotes-gw.ent.nwie.net>



We are running a "shared server" environment with Sun 490s in active/passive configuration with Veritas Cluster. In each cluster we have four nodes - two (primary) db nodes, one application node, and a passive node for failovers. We have a couple of different setups - a "class 1" that has global failover (different data centers), and a "class 2" that can only fail over within the cluster in one data center.

The thing we are planning for is hardware failures. If a board goes bad or we lose a whole server, we can fail over to the passive node until the primary is fixed. The same applies for maintenance. We also chop things up into Service/Resource groups, so a particular suite of databases (say, one application uses dev/test, uat, and pt databases) resides in one service group. The disks and such (disk group, mounts, oracle_home, listener, etc.) are put in only one service group, and it can fail to different nodes for maintenance, load balancing, etc. if needed. This also allows for something like a disk failure to only impact the service group that contains that particular hardware resource, and it can be failed over to a different node while the service groups & databases that don't use that resource can continue to run on the primary node. The prod databases are split up in a similar fashion.

From:
Ken Simpson <ipadba_at_gmail.com>
To:
oracle-l_at_freelists.org
Date:
01/14/2010 02:35 PM
Subject:
Server consolidation and minimizing unplanned downtime Sent by:
oracle-l-bounce_at_freelists.org

Over the past couple of years our hardware group were real big on server consolidation so we ended up with a pair of Sun boxes supporting between 15 and 20 databases each. The SLAs didn't warrant any type of HA requirement so they are standalone boxes and in the event of a hardware failure the business was okay with waiting for Sun to service the gear. The wonderful world of shared services and being able to pull data from anywhere at anytime has meant that we now need some sort of redundancy in this environment. All of the databases are small (between 10GB and 150GB in size). The majority of these database support 3rd party packages from small vendors who are historically very slow with supporting recent releases of Oracle so we're forced with supporting everything back to 9i.

For those of you with environments containing 30+ databases with releases running anything between 9i and 11g, how are you providing redundancy to minimize unplanned downtime? We don't need 5 9's availability and can schedule the occasional maintenance. We just need a way to minimize downtime ( < 15 minutes ) in the event of an unplanned outage or hardware failure.

RAC? Veritas (or Sun) Cluster? Golden Gate?

TIA
--

http://www.freelists.org/webpage/oracle-l

--

http://www.freelists.org/webpage/oracle-l Received on Thu Jan 14 2010 - 13:50:17 CST

Original text of this message