Comparing Kernel.

[joe_91]

Hi:

I am comparing kernels on two boxes (one test and one Production) and find the following differences...

maxswapchunk 4096(test) 4092(prod)
maxvgs 10(test) 30(Prod)

Also as mentioned in the discussion below the Prod. does not have the English HP-UX 64-bit Runtime Environment, while the test has got that bundle installed. Will the above parameters make a difference????

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xc3265fe8b250d71190080090279cd0f9,00.html

Please Help!!

Thanks

Joe.

[Ken Hubnik_2]

No unless you are going to use more than 10 volume groups on test

[Steve Steel]

Hi

maxswapchunks is used to define the maximum amount of swap and since the difference is minimal .no difference.

maxvgs is the number of volume groups on a machine and this definately does nothing.

see
http://docs.hp.com/hpux/onlinedocs/TKP-90202/TKP-90202.html


Steve steel

[A. Clay Stephenson]

The differences in maxswapchunk are so small as to make little difference. The one difference would be that "test" supports a few additional MB's of swap - but this is a limit. Chances are that you have not allocated that much swap. Similarly, maxvgs - maximum number of VG's allowed should be ok (unless more VG's are needed). Still, whenever possible (or unless intentionally not done), I try to make the sandbox and test environments mirror the production. There are simply fewer surprises that way. For example, 10 may be plenty of VG's for "test" but "30" might be too small for production. It depends upon how closely you endeavor to mimic the two environments.

[Michael Steele_2]

maxswapchunk as 4092 should be an octel number like, 4096. So this looks like a minor typo.

maxvgs is just what it suggests, a ceiling for the number of vg's on your server. Do you have more than 10?

[S.K. Chan]

No they wouldn't.
1) maxswapchunk is use to determine the maximum amount of swap space you can configure on your system.
2) maxvgs defines the maximum volume groups allowed on your system.

[Bill Hassell]

There's nothing mysterious about these parameters:

maxswapchunk is the size of the swap space allocation unit and if it is too low, you can't add any more swap space. Making it larger (like 8192, etc) has virtually no effect on the operations but allows for several Gb of swap space to be defined.

maxvgs 10(test) 30(Prod)

maxvgs is simply the limit for the number of Volume Groups allowed in LVM. If you need more than 10 VG's in test or more than 30 in prod, you'll have to change these values as needed.

If you are comparing the two kernels because of problems during operation (ie, performance), these parameters have no effect other than mentioned above.

[joe_91]

Thanks. But what about the English 64 bit runtime Environment that is installed on the test and not on the production???

Thanks
Joe

[A. Clay Stephenson · Accepted solution]

The 64-bit OE is more serious - at least potentially. As long as the test/development box is running 32-bit, I would not be concerned with moving to a 64-bit production environment BUT the converse is not true.
Don't be deceived by commands which simply indicate that the CPU will support 64-bit. That does not mean that the kernel and other OS utilities are 64-bit.

The trickiest mixed 32-bit/64-bit applications that I've found involved sharing memory between a segment created by 64-bit application but accessed by 32-bit applications. It can be done but the developer must be a disciplined individual.

[joe_91]

Thanks Clay. But if you would expand on your statement a little more that would really help.

Thanks

Joe.

[A. Clay Stephenson]

Hi again Joe:

I couldn't open your previous thread - the Forums seem to be crawling - but here's my bottom line. If your test/development box is 32-bit
and your production box is 64-bit then I would expect absolutely no surprises (unless the application is kernel intrusive e.g. lsof). If on the other hand, the development/test box is 64-bit and the production is 32-bit then I would feel VERY much less confident. My approach would be to do an OS upgrade (yes, upgrade to 64-bit - it does work). Of course, first do an Ignite. The upgrade works if you intentionally oversized / and /stand a bit.

I suspect that (without knowing the models of the two boxes) is that one was installed 32-bit and the other 64-bit. It's easy to
do.

You should also make certain that you choose the same applications (e.g. Oracle can be loaded 64-bit or 32-bit) "bitness" as well.

Without meaning to sound harsh, smug, --- all the other Warm Fuzzy stuff, making sure that the test and production environments match as closely as possible (if that was intended) is something that should have been addressed much earlier in the process.

[joe_91]

Hi Clay/Team:

Here is the difference on the test and the production. (HP-UX 64-bit Runtime Environment) Both machines say 64 bit on getconf and what /stand/vmunix. Please advise.

Thanks
Joe


# Initializing...
# Contacting target "test"...
#
# Target: test:/
#

A5506B B.11.00.02 PCI 4 PORT 100BASE-T/9000
A6795A B.11.00.10 PCI Tachyon TL/TS/XL2 Fibre Channel
B2491BA B.11.00 MirrorDisk/UX
B3701AA C.03.70.00 HP GlancePlus/UX Pak for s800 11.0
B3929BA B.11.00 HP OnLineJFS (Advanced VxFS)
B3935DA A.11.14 MC / Service Guard
B5139DA B.01.09 Enterprise Cluster Master Toolkit
B5725AA B.4.1.55 HP-UX Installation Utilities (Ignite-UX)
B7609BA A.03.20 Event Monitoring Service
B8324BA B.01.04 HP Cluster Object Manager
B8342AA B.11.00.10 Netscape Communicator 4.79
B8724AA A.01.08 CIFS/9000 Client
B8725AA A.01.08 CIFS/9000 Server
CMDVIEWSDM A.1.05.00 hp StorageWorks CommandView SDM
HPUXEng64RT B.11.00 English HP-UX 64-bit Runtime Environment
HWE1100 B.11.00.0212.4 Hardware Enablement Patches for HP-UX 11.00, December 2002
OnlineDiag B.11.00.23.10 HPUX 11.0 Support Tools Bundle, Dec 2002
QPK1100 B.11.00.58.5 Quality Pack for HP-UX 11.00, September 2002
UnlimUserLic B.11.00.02 HP-UX Unlimited-User License
perl B.5.6.1.C Perl Programming Language


# Initializing...
# Contacting target "production"...
#
# Target: production:/
#

A5506B B.11.00.02 PCI 4 PORT 100BASE-T/9000
A6795A B.11.00.10 PCI Tachyon TL/TS/XL2 Fibre Channel
B2491BA B.11.00 MirrorDisk/UX
B3701AA C.03.70.00 HP GlancePlus/UX Pak for s800 11.0
B3929BA B.11.00 HP OnLineJFS (Advanced VxFS)
B3935DA A.11.14 MC / Service Guard
B5139DA B.01.09 Enterprise Cluster Master Toolkit
B5725AA B.4.1.55 HP-UX Installation Utilities (Ignite-UX)
B7609BA A.03.20 Event Monitoring Service
B8324BA B.01.04 HP Cluster Object Manager
B8342AA B.11.00.10 Netscape Communicator 4.79
B8724AA A.01.08 CIFS/9000 Client
B8725AA A.01.08 CIFS/9000 Server
CMDVIEWSDM A.1.06.00 hp StorageWorks Command View SDM
HWE1100 B.11.00.0212.4 Hardware Enablement Patches for HP-UX 11.00, December 2002
OnlineDiag B.11.00.23.10 HPUX 11.0 Support Tools Bundle, Dec 2002
QPK1100 B.11.00.58.5 Quality Pack for HP-UX 11.00, September 2002
UnlimUserLic B.11.00.02 HP-UX Unlimited-User License
perl B.5.6.1.C Perl Programming Language

[joe_91]

Team:


Any more thoughts on 64 bit-runtime???

Thanks

Joe.

[Steven E. Protter]

I don't think those differences or the discussion in the other thread will make much difference.

maxswapchunk 4096(test) 4092(prod)
maxvgs 10(test) 30(Prod)


The swapchunk figures are not statistically significant as far as their difference.

If maxvgs is what I think it is, its not likely to matter either. I have two, max three volume groups on my most complex of servers.

SEP