General question about frequency of Kernel corruption with HP-UX 11i.

[Joanne Keegan]

This is just a general query to see if anyone else is noticing the same thing or not.

Most of our servers are now running HP-UX11i. When installing patches or drivers, there seems to be a higher frequency of a corrupted kernel than we ever experienced on HP-UX 11.00.

The procedures we are using are correct, yet the frequency of corruption is becoming a worry.

Anyone else noticed this?

Regards,

Joanne

[A. Clay Stephenson]

I haven't noticed this at all although when you say 11i that might not mean solely PA-RISC. If by 11i, you mean 11.11 then I have seen no increased rate of "corruption". It would help if you defined that term a little better. I suppose you mean that patches were loaded that left you with a kernel that could not be built.

The most frequent cause of this is failing to load all the required patches. I tend to not do custom patch installs but install the entire Gold QPK's first on a Sandbox, next on Test, and finally onto Production. I haven't had a bad patch hairday in several years following that policy.

[Steven E. Protter]

The only time I noticed this on 11.00 was when PHKL_18543 got installed twice by me no the same system.

Besides that nasty episide I have nearly a decade of experience totallying up machine run time on 11.00 without such issues.

I'd check into the PHKL_18543 issue and other possible human causes.

I found 11.00 to be reliable and stable and other than what I mentioned above not corrupted.

SEP

[Patrick Wallek]

I have not seen any kernel problems on 11.11 after installing patches. I have to agree with Clay in that the culprit is usually not installing ALL required dependencies. The dependencies are an absolute MUST. They are not optional.

Fortunately the ITRC Patch tool has gotten much better about selecting all dependent patches when you select a patch to download. I have not had a single problem with this yet.

[Joanne Keegan]

The version we are running is HP-UX 11.11 on PA-RISC servers.

[A. Clay Stephenson]

One thing that occurs to me is that you might have altered /var/ad/sw/defaults (or used a -X specified config file) that sets enforce_dependencies=false. That could certainly cause problems. Also, if some of your filesystems are pressed for space, that too, could cause problems. After all, swinstall's size requirement estimates are just that.

[Steven E. Protter]

Missed the question on first hack.

Absolute no problems with the kernel in 11i(11.11).

Not one problem in nearly six machine years (2 boxes two years old this May).

11i was substantially more reliable than 11.00 and substantially harder to break.

SEP

[Sanjiv Sharma_1]

Hi Joanne,

I have never faced for the last 2 years any problem with the patch installation in HP-UX 11i (11.11).

We have applied various patches as and when required in almost 15 servers w/o any problem.

Can you be more specific on the error recieved or the actual problem observed?

[Bill Hassell · Accepted solution]

Like other posters, I am not clear about kernel corruption. To me, kernel corruption is the modification data pointers and tables internal to the kernel, but the reason it occurs would be due to mistakes in kernel code, almost always fixed by patches.

Now if you mean that after installing patches and rebuilding the kernel that the kernel build process fails, that's different. In that case, the problem is with the integrity of the libraries used to build the kernel and for recreating the /stand/system file, the current /stand/vmunix file. For a production server, I would not start patching are changing kernel params or drivers without running a quick kernel generation. If it works OK, I'll continue [patching, otherwise, work on fixing the kernel build steps.

[Dietmar Konermann]

Joanne,

I tend to agree with the others. The vast majority of cases where kernel-rebuilding fails happen due of missing patch dependencies (and due to PHKL_18543 re-installations, of course :-)).

The situation on >=11.11 should be even better since SD-UX has been enhanced to fully support the "corequisites" tag for patches. This avoids lots of problems... unless you use "enforce_dependencies=false".

Best regards...
Dietmar.