Previous | Table of Contents | Next |
Perhaps this is another legacy of my submarine days, but I always like to have a plan in place to handle the situation wherein problems come up and I have to abandon my upgrade efforts. Perhaps it is just that I do not want to have to wing it when the problems come up and everyone is yelling for a solution, or perhaps I just want to assure my customer or boss that I can handle it if the software is bad or some other problem arises. This is especially important when performing upgrades to production systems.
What are some of the items that should be in this backout plan? Well, the first thing is the criteria that would cause you to have to perform the backout. Its a good idea to get management people to agree to this in advance so that they do not act like armchair quarterbacks the next day and tell you that you should have done it sooner or waited a little longer. State the criteria in terms of time (If there are still problems at 5 p.m. and there is no immediate solution, begin backout so the system can be operational for nightly batch processing) and other conditions (There is a bug that Oracle needs to investigate and it indicates it may be several days). Again, much like the installation plan, send the backout plan to managers (project managers, DBA group managers, and so forth) so that they can look at it and get their shots in before problems occur. Make them understand that if they do not give you comments, you have their approval to proceed according to the plan.
The main part of the backout plan is the steps needed to recover the instance fully so that it can run under the old software. This usually means restoring a complete, cold backup of the data and resetting parameter files, such as oratab, to point to the old version of the software. List all the files that need to be restored. This serves two purposes. First, you do not have to rely on your memory when it is late, you have been struggling for hours with a particularly annoying bug, and you have to restore the old instance to get production going again. Second, it serves as a good checklist for your backup that is performed as part of the upgrade plan. You really want to ensure that you have backed up all the files that your backout plan will require. This may seem simple on small, single-disk computer systems, but when you have files scattered across 135G of disk or more, it can be a challenge. Remember, Oracle is not forgiving if you remember to restore only 95 percent of the files that it needs.
As has been mentioned frequently in this chapter and Chapter 6, there can be problems with the installation and upgrade process. They can be worked through, but it takes time and support from other people to make the installation and upgrade process work. Specifically, there are a few resources that you should consider lining up during the planning stage so that you can have them ready when you actually perform the upgrade:
Previous | Table of Contents | Next |