10 years ago

Rolling Database Upgrade: The UPS Experience

Share in:
Share in:

 UPS was ready to test the rolling database using their own system.

UPS is the world leader in transportation, starting out as a small company that provided messenger services to a top provider of ocean, ground, electronics and air services.  UPS is currently one of the biggest airlines the world over and its WorldportSM is the biggest automatic package handling facility in the world.

WorldportSM serves close to 800 airports across the globe.  And with more than twenty years operating as an airline, UPS needed to upgrade their databases to ensure that they are able to handle the future.

Rolling Database Upgrade

The company opted to have a rolling database upgrade because they could not afford to undergo the downtime that is normally expected with a conventional database upgrade.  Also, they needed a way to test the upgrade process first in laboratory or development environment before integrating and certifying the upgrade process in an application test environment that approximates real-world situations and load.  The company also wanted to document the whole upgrade process to their existing database upgrade strategies.

Even though the upgrade constitutes a planned down time, the service level agreement limited the maximum allowed downtime for a single event to only 15 minutes.  Before Oracle 11g and the ability to have rolling upgrades, that was simply not possible.  In fact, upgrading from an earlier version to 11g takes about 55 minutes of downtime.  Thanks to rolling upgrades, you can be done with downtime in only 4 minutes.  That means you get an extra 11 minutes out of your SLA!

UPS also set up a list of critical tasks failures and corresponding fallback options.  For example, if an upgrade of a logical standby database fails, you have the option to flashback the upgrade or to downgrade.  On top of that, they also set up a list of database and site failures, and corresponding fallback options using transient logical.

After that, UPS was ready to test the rolling database using their own system, which includes their flight operations and scheduling application, application servers and database servers.

The company found that they are able to test a rolling upgrade with a real world application, and also change the application’s database connections, including changes to the database triggers and database services.  They could even use the same application servers for the switchover of the databases and test the process even with the application still running.  All in all, UPS was able to do the database upgrades with less than four minutes of downtime.

Enter Transient Logical Standby

UPS also checked the rolling upgrade and how it affected Data Guard and streams.  They encountered some problems with the stream metadata because of unsupported SQL APPLY data types.  This was solved using another Oracle feature: Transient Logical Standby.

Overall, rolling database upgrades benefited UPS by allowing them to configure a company application to be able to reconnect to the database when there is an application switchover, even while the application is still running.  And that you can use transient logical standby in rolling database upgrades to smooth out streams.  And most importantly, downtime was kept to a minimum.

So if you are growing to a point that you need to upgrade your database, call Four Cornerstone today.  We can help you do a rolling database upgrade without you needing to take your critical applications offline too long.

Source: Oracle.

Scroll to Top