Building a Bulletproof Penguin: DB2 and Linux Failover
Joe La Chapell and JT Vogt
Implementing a new enterprise
resource planning (ERP) system that will impact every division of a company is an
intriguing proposition. A lot is expected of such a system, and when we were tasked
with this undertaking at our company, we did not want to neglect the foundation
of it -- the database. As we researched this project, it became evident that
we could satisfy all the necessary requirements with a combination of products:
IBM DB2, Red Hat Linux, as well as Heartbeat, Mon, and ipfail. In this article,
I will describe how we built a "bulletproof penguin" one piece at a time.
No Single Point of Failure
We needed a solution that would guarantee high
availability and failover, so we decided to install two database servers implementing
DB2's new High Availability Disaster Recovery (HADR). We also wanted to make
certain we had no single point of failure in our hardware or software. We purchased
a 14-disk storage array (SAN) and allocated one large RAID 5 partition for the database
servers to use for storage. Our SAN controller has no single point of failure, so
from a hardware perspective, we are on solid footing. Our HADR solution provides
no single point of failure from a software perspective. Since we had committed to
DB2, we had to use a shared nothing approach, which, simply stated, means that each
node in the cluster has its own file system.
Our Disks
Once we had our disk array, we were able to install
our operating system on each of the database servers. We wanted an operating system
that was scalable and reliable, so selecting Red Hat Advanced Server as our operating
system was a pretty easy choice.
|
