[pca] Recommended patches and PCA

Martin Paul martin at par.univie.ac.at
Tue Oct 18 14:33:48 CEST 2011


Hi Jonathan,

> However I am a little confused about the way PCA deals with Recommended
> patches.

This can indeed be a little confusing. For a long time, only the latest revision 
of each patch was included in patchdiag.xref, so that's what PCA was made to 
work with. There were different patch clusters, and the "R" flag in the xref 
file didn't exactly match any of them. So PCA defined its "recommended" set of 
patches to be the latest revision of every patch which is marked "R".

Last year, Oracle changed that. There is now only one "Recommended Patchset for 
Solaris". It contains the lowest revision of all patches which match certain 
criteria (recommended, security, etc.). Details are at:

   http://blogs.oracle.com/patch/entry/a_solaris_recommended_patchset_to

The usage of the "R" flag in the xref file was modified, too. If a recommended 
patch gets a new revision which does not include a new recommended fix, the 
recommended revision is kept in the xref file marked as "R" and "O" (obsolete), 
as seen in your example:

> 118712|23|Mar/19/09|R| |O|  |10|
> 118712|24|Aug/06/10| | | |  |10|

So now I had to decide how to handle this new situation in PCA. For some people, 
it makes much more sense to install 118712-24 instead of 118712-23 in a 
situation that you describe: If I take the risk of installing a patch, I can 
just as well install the most recent revision instead of an outdated one. I 
therefore modified PCA to promote the "R" flag to the latest revision.

Still, there was interest in using PCA to install exactly the same set of 
patches as in the new Recommended Patchset. I therefore added a new option 
"--minimal" to PCA, which can be used in combination with the "missingr" patch 
group. It will restrict itself to the last revision which actually carries the 
"R" flag, ignoring newer non-"R" revisions.

As I don't use "--minimal" in production, and never received much feedback, I 
cannot guarantee it to work 100% correctly: All my tests were fine, but the real 
world (and changes in the Recommended Patchset) might be a different thing. So 
the option was never documented officially, but you can feel free to use it (and 
are welcome to report your experiences here).

If you already download the "Recommended Patchset" anyway, you should of course 
ask yourself whether it wouldn't make more sense to use the included patch 
install script instead of PCA? The only real advantage of "pca --minimal 
--install missingr" is probably the (greatly) reduced amount of patches to be 
donwloaded from Oracle. Otherwise, the end result should be the same.

I hope I could shed some light on this issue and it helps you to decide on how 
to go on!

Martin.



More information about the pca mailing list