Increasing the size of the shared memory that can be locjed by sybase on hpux11.00

[Osunkoya Olugbenga Davi]

I have tried increasing the shmmax parameter to 4GB in the kernel parameter but did not resolve the problem of allowing Sybase to use a shared memory size of 2.7GB.

I have been able to achive 2.0GB by tuning some Kernel parameters. How do I go around this?

Thank you.

[Mobeen_1]

Osunkoya,
We have done this many a times on Solaris and Sybase. Its been quite some time and i don't remember the precise command, but could tell you how we would achieve it

1. Make sure that 4GB of shared mem can
be seen at your OS level

2. Then in Sybase use sp_configure and
increase the parameters for shared memory
I know these were not dynamic.

3. Once that is done, shutdown your Sybase
ASE and start it up

regards
Mobeen

[Osunkoya Olugbenga Davi]

I have increased my SHMMAX size to 6GB, but could not start the sybase sybase service with more than 2G in the memory allocation in the sybase config file. See attached report generated when starting with 2.7GB memory in sybase.

What could be the next step?

Thank you.

[Mobeen_1]

Osunkoya,
From what i see from the error, the first issue we would need to handle is

1. Do you have a .cfg file specified in your runserver file namely RUN_ABUJA ?

If you have specified, make sure that the location is right ?

Also please post that .cfg file here, so that we can make sure that its ok.

Please do also make sure that the HP-UX host has been rebooted after increasing the shmmax param and that at the OS level the shared memory is what you expect

rgds
Mobeen

[Mobeen_1]

Also from your attachment, i can see that your default config file at /opt/sybase/ABUJA.cfg is trying to allocate only 748052480 bytes (approx 700 odd MB).

rgds
Mobeen

[Osunkoya Olugbenga Davi]

Please find attached the ABUJA.cfg file for the 2.75GB.

The server has been rebooted after the SHMMAX was chacged to 6GB.

What di I need to next?

Thank you

[Robert-Jan Goossens]

Hi,

Are you running 32 or a 64 bits system ?

# file /stand/vmunix
vmunix: ELF-64 executable object file - PA-RISC 2.0 (LP64)

Robert-Jan

[Steven E. Protter]

kernel ignire shmmax settings greater than 25% of memory.

memory is defined in this case as ram plus swap.

So, pumping up shmmax is not going to do any good beyond that situation.

To see, swapinfo -tam

You can add physical memory, this will make more available to shmmax. You can add swap, this does the same but has a significant, negative performance impact.

SEP

[Osunkoya Olugbenga Davi]

Please find the result of the file /stand/vmunix and swapinfo -tam.

ab_s_03a:/ # file /stand/vmunix
/stand/vmunix: ELF-64 executable object file - PA-RISC 2.0 (LP64)

ab_s_03a:/ # swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1024 0 1024 0% 0 - 1 /dev/vg00/lvol2
dev 4096 0 4096 0% 0 - 2 /dev/vg00/lvol10
reserve - 526 -526
total 5120 526 4594 10% - 0 -
ab_s_03a:/ #

Please, remember that I have 11Gb of physical memory on the server.

Also, below is the list of patches that I installed on hpux11.00 before this problem started for the installion of c8xx contoller cards for DS2300 disk system;

PHC0_21187
PHCO_23651
PHKL_18543
PHKL_20016
PHKL_22588
PHKL_24279
PHKL_27980
PHKL_28766
PHKL_29256
PHKL_29494