Clustering for Operational Efficiencies

[M__Braun]

Normal 0 false false false MicrosoftInternetExplorer4 While it’s true that clustering is targeted primarily atproviding high availability, advanced clustering solutions are not just about HA. They can also make lifeeasier for IT administrators.

Think about it. Clustering enables you to automaticallystart and stop applications within a cluster. If you put a lot of applicationsthat are across a lot of servers into a single cluster, then you also have avery easy way to move applications around all those servers.

Let’s say Joe Operator is sitting in the NOC at night when hegets a call saying Oracle Instance 1 needs to be moved from System 1 over toSystem 8 because the team needs to do maintenance. Joe could “su” over to theOracle user account, issue a server manager control stop command to halt thatOracle instance, unmount file systems, bring down IP addresses, and then go toanother server and reverse that process.

Impressive.

Or, Joe can look at his cluster management GUI, right-clickto get the particular application he wants, tell it he wants to move it fromSystem 1, where it is currently running, over to System 8, which is currentlyan idle system in the cluster, and be done with it. With a few mouse clicks,the application has moved. Joe didn’t even need to have root or administratoraccess to the servers to make the changes; all he needed was access to thecluster management GUI.

Infinitely more impressive.

So, what other value does Joe get from this? For starters,he isn’t relegated to doing kernel patches or OS upgrades in the middle of thenight. He can instead apply the upgrades to a spare server in a cluster, leavea change request with the NOC asking them to move the application over to theupgraded server at a convenient time during off-hours, then come in thefollowing day to an upgraded node, complete with application running, and to anew idle node … which he can then use for the same or similar purposes.

Because Veritas Cluster Server supports multiple versions ofan operating system within the same cluster, it’s easy to do kernel patches, OSupgrades, and hardware maintenance without affecting any running applications.So, having a passive spare server not only allows for rolling applicationupgrades but it also can be used to perform OS upgrades or patches. After afull test of the first upgrade, the next upgrade can be done. What’s more, abuilt-in failback path ensures that if the first upgrade is unsuccessful, then itwill automatically switch back to the original.

And Joe is free to have a good evening.