[pca] synchronization between "recommended patch cluster" andpatchdiag.xref?

[Paul B. Henson]

On Wed, 21 Oct 2009, Ben Taylor wrote:

> Did you send them an explorer?

Yes, it was based on that that they were complaining about missing patches.

> If you have NFS problems, then it's no wonder they're saying you're out
> of date.

I didn't mind them saying we might want to try these patches, but it came
along with an implication that we should be installing the recommended
patch cluster and would have already have them. Which it turns out wasn't
the case.

> I usually push back if some solution center yo-yo tries to tell me I
> should be updating patches that have nothing to do with the issue.
> That's the solution center's basic mantra.  Update, update, update.
> Which I can't fault them for.  It those in-duh-viduals which insist that
> some patch be installed which has no correlation to the problem at hand
> which raises my ire.

Yes, it is indeed infuriating to get a zombie "must apply patches" response
from support. We generally install almost all recommended patches, most
security patches, and a smattering of other patches as they seem applicable
to our environment. I review the details of all the patches released, which
is somewhat time-consuming but generally gives me a good handle on what's
going on. It takes time and resources to apply patches, so unless support
can provide a specific bug that a particular patch fixes which seems like
it might involve our problem we often push back, particularly if the patch
requires downtime.

> With ZFS root, I'm just installing patches into a boot environment.  I
> don't have the cycles to match against recommended, and so I just let pca
> update to current.

You install everything without reviewing what's changed? I don't think I'd
want to do that. Another reason not to blindly install patches is that more
frequently than I would like either a regression or a completely new bug is
introduced in a new patch :(. I'm not sure what you mean about cycles to
install only recommended, it seems adding an 'r' to 'missing' would take
care of that for you.

Also, because of a horrible lack of scalability with sharing zfs
filesystems over NFS (which I have had an open case on for well over a
year), it takes a bit over two hours to completely reboot our x4500 servers
with about 8000 filesystems each, which makes it rather difficult to
schedule downtime for non-essential patching.


-- 
Paul B. Henson  |  (909) 979-6361  |  http://www.csupomona.edu/~henson/
Operating Systems and Network Analyst  |  henson at csupomona.edu
California State Polytechnic University  |  Pomona CA 91768