The risks of Standard Patch Bundles installation

[manamirhastam]

We want to apply the latest version of Standard Patch Bundles(Jun '07) on rx servers including HP-UX 11.23
Is there any risk for applying the mentioned bundles on a live environment?
Such as any change on Kernel parameters values or something else like this.

[Pete Randall]

That is exactly why the bundles are created: to minimize the risk. The patches in the bundle have been tested, together, to ensure that any risk is absolutely minimized.


Pete

[James R. Ferguson]

Hi:

The standard patch bundles to which you refer are the most generally applicable, most throughly tested sets offered.

As with any patch session, you are advised to create an appropriate Ignite recovery (tape) beforehand AND read the patch notes for any special installation instructions or warnings.

A reboot will be required to apply the bundle, so plan accordingly. You should apply the bundles when the system is in a normal operating mode --- networking up as that is required for 'swinstall' to operate properly.

A new kernel will be generated since the patch bundles do contain PHKL and PHNE (kernel and networking) patchsets.

Regards!

...JRF...

[A. Clay Stephenson · Accepted solution]

Except in extremely rare cases are kernel tunables ever changed by a patch. What is much more common is that a new tunable is added and assigned a default value that might cause some unexpected change in behavior. The risks associated with applying a patch bundle are never zero. There is simply no way to test every combination and permutation of installed hardware and software. However, in almost all cases the risks of NOT applying a patch bundle greatly exceed the risks of applying the same bundle.
Remember, these patches were created mainly to resolve known problems.

In the vast majority of cases, the testing of Standard Patch Bundles is thorough so that the risks are minimal. Some shops prefer to wait a few months so that someone else will have already experienced any remaining unexpected problems. However, that approach has risks as well. I always look for patches that mention things like "possible data corruption". Those should be carefully studied and generally applied before the standard bundles are available. There is no worse problem that a bug that can trigger minor changes in data scattered all over a disk. You might not even know there problem exists for weeks and by then there is really no practical recovery. Fortunately, those types of bugs are very rare. Another approach to take is to do an analysis and install a customized bundle but as a general rule the standard bundles work well and are a reasonable approach.

My approach to the standard bundles is three-tiered. 1) As soon as the bundle is available, I deploy it in my sandbox environment and wait a few days. If it explodes I really don't care. 2) I deploy it in the test environment and wait at least two weeks. 3) After requesting a maintenance window, I deploy in production.

Using this approach, very few bugs survive to the production environment.

With respect to the June '07 Bundles; I've deployed them and had no problems BUT that does not mean that you will not have any --- it's the nature of the beast.

[manamirhastam]

Thanks all