Poor performance after upgrading an 11.00 box to 11i..

[Richard E. Flower]

We've got a 9000/800 2 processor server that was recently running HPUX 11.00 without a hitch.. In that configuration it had good response time,etc. We had it upgraded to 11i, and ever since it's really been a dog performance wise.. It takes very little to tie up the entire machine in knots.. I recently installed a bunch of patches (see below) in the hopes that it would solve some of the issues we've been having.. Unfortunetly, no such luck.. Any ideas?

#
# Bundle(s):
#

B3901BA B.11.11.01 HP C/ANSI C Developer's Bundle for HP-UX 11.00 (S800)
B3913DB C.03.33.01 HP aC++ Compiler (S800)
B5725AA B.3.2.45 HP-UX Installation Utilities (Ignite-UX)
BUNDLE11i B.11.11.0102.2 Required Patch Bundle for HP-UX 11i, February 2001
CDE-English B.11.11 English CDE Environment
FDDI-00 B.11.11.01 PCI FDDI;Supptd HW=A3739A/A3739B;SW=J3626AA
FibrChanl-00 B.11.11.03 PCI FibreChannel;Supptd HW=A5158A
GigEther-00 B.11.11.14 PCI/HSC GigEther;Supptd HW=A4926A/A4929A/A4924A/A4925A;SW=J1642AA
HPUX11i-OE B.11.11 HP-UX Internet Operating Environment Component
HPUXBase64 B.11.11 HP-UX 64-bit Base OS
HPUXBaseAux B.11.11 HP-UX Base OS Auxiliary
Ignite-UX-10-20 B.3.2.45 HP-UX Installation Utilities for Installing 10.20 Systems
Ignite-UX-11-00 B.3.2.45 HP-UX Installation Utilities for Installing 11.00 Systems
Ignite-UX-11-11 B.3.2.45 HP-UX Installation Utilities for Installing 11.11 Systems
OnlineDiag B.11.11.00.04 HPUX 11.11 Support Tools Bundle
RAID-00 B.11.11.00 PCI RAID; Supptd HW=A5856A
#
# Product(s) not contained in a Bundle:
#

C-ANSI-C B.11.11.02 HP C/ANSI C Compiler
C-Analysis-Tools B.11.11.02 C Language Analysis Tools
DebugPrg B.11.11.05 Debugging Support Tools
PHCO_23251 1.0 libc manpage cumulative patch
PHCO_23463 1.0 sysdef(1) patch
PHCO_24402 1.0 libc cumulative header file patch
PHCO_24716 1.0 tar(1), checksum, special character
PHCO_25226 1.0 Initialised TLS, Psets, Mutex performance
PHCO_25452 1.0 libc cumulative patch
PHCO_25568 1.0 libnss_files cumulative patch
PHKL_23242 1.0 VxFS mmap(2) performance improvement
PHKL_23374 1.0 Fix close(2) data corruption
PHKL_25233 1.0 select(2) and poll(2) hang
PHKL_25389 1.0 Core PM, vPar, Psets Cumulative Patch
PHNE_25084 1.0 Cmulative STREAMS Patch
PHNE_25642 1.0 cumulative ARPA Transport patch
PHSS_24304 1.0 ld(1) and linker tools cumulative patch
PHSS_24638 1.0 HP aC++ -AA runtime libraries (aCC A.03.33)
WDB B.11.11.05 HP Wildebeest (HP WDB) Debugger
WDB-GUI B.11.11.05 GUI for the HP WDB Debugger

[Sanjay_6]

Hi Richard,

suggest you look at the kernel parameters,

Try this guide,

http://docs.hp.com/hpux/onlinedocs/os/11.0/tuningwp.html

And here is the list of kernel parameters,

http://docs.hp.com//hpux/onlinedocs/os/KCparams.OverviewAll.html

Do a search on forum with the keyword as performance and you'll find a lot many suggestion which you can try to improve your system performance.

Hope this helps.

Regds

[James R. Ferguson]

Hi Richard:

What does 'glance' show? Are your kernel parameters the "same" as on 11.0 -- particularly 'timeslice' and 'max_dbc_pct' (for buffer cache)? More specific details of performance metrics are needed.

Regards!

...JRF...

[A. Clay Stephenson]

Hi Richard:

Displaying the swlist is of very limited value. You need to do some digging with top,
sar, ps -ef, etc. You should also post your kntune output. The best tool to use would be the Trial version of Glance from any of your Application CD sets.

The only performance killer that I've seen with 11.11 was a problem with the early On-line Diagnostics. One of the deamons was a CPU hog but if your have installed any of the General Release patchsets sine June 2001 that has been fixed. There is just not enough data here to begin.

[Patrick Wallek]

What specific type of machine? All servers are 9000/800.
How much RAM?
How much swap space?

Have you had a look at glance or top when the machine slows down? If you don't have glance you can install a trial version from the Application CDs, I believe CD #1.

[A. Clay Stephenson]

The first thing I would do is to lower dbc_max_pct to no more that about 20% and 15% is probably better. You have a huge amount of swap and a small amount of memory and I am seeing some page outs. I would set the value of swapmem_on to zero. That should only be set to one if you have much more memory than swap which is not your case.

[Sanjay_6]

Hi Richard,

Configrue your dbc_mac_pct to 10%. This should help in improving the performance.

hope this helps.

Regds

[Richard E. Flower]

Thanks guys for the help! I'll give it a spin down to 5 and see how things behave.. Let's see if I can dump everyone off first! (:->

[Richard E. Flower]

Just a quick followup.. I cranked the dbc_max_pct down from 50 to 10 and the machine appears to be noticably more spunky.. Thanks all for the help!

-- Rick