memory performance question...

[James Ellis_1]

We have a server that has three databases. One of the database cannot startup while the other database is up and running. DBA seems to think its a memory problem.

swapinfo shows a 1G memory, but systems properties under SAM show a 8G memory. swapinfo also shows PCT USED as 0%. Even with one DB up, the PCT shows 0%. This does not appear right. Would ipcs be able to show the total memory used by all of the processes?

Right now, we are able to start one DB, but not the other two. Something is preventing these DB from coming up, and I'd like to verify this is not a mamory problem. Any help?

Thanks.

[Ian Dennison_1]

What are the symptoms that you get when trying to start the databases? Does /var/adm/syslog/syslog.log or 'dmesg' show anything that looks like it ties in with the dates / times you start the database?

Also, look at 'vmstat' for File tables (specifically nproc and maxuprc).

Share and Enjoy! Ian

[Jeff Schussele]

Hi James,

Sounds to me like the DBs are 32-bit & are all trying to run in the default memory window.
Remember that 32-bit apps are limited to 4 1GB quadrants with all the data extents sharing the same 1GB quad.
If this is the case what you need to do is implement memory windows.
Search the forum or docs.hp.com for "memory windows" for further info on how to implement & use memory windows.
Also it's possible that you are painfully short on swap space - run
swapinfo -tam
and check just how much swap space is available for reservation. ALL processes must be able to reserve swap space - not use it - just put a placeholder there in case it needs to use it. You *must* have swap space equal in memory space to all processes currently running or wanting to run.

Rgds,
Jeff

[Massimo Bianchi]

hi,
first you must specify which kind of DB: oracle, informix, postgres, mysql; then the compiled version: 32 or 64 bits.

Finally your hpux version: 11.00, 11.23, and 32 or 64 bits.


Otherwise, there might be far too problem in guessing , than in the real problem.

Regards,
Massimo

[Massimo Bianchi]

HI,
among the other informarion: the /stand/system config file.

Maybe it's a simple issue of semaphores or shared memory segment number...

HTH,
Massimo

[James Ellis_1]

We are running Oracle
DBs are 64 bit
System is HP-UX 11.0

I am still looking into this...

Thanks.

[James Ellis_1]

Jeff,

You stated this:
Also it's possible that you are painfully short on swap space - run
swapinfo -tam
and check just how much swap space is available for reservation. ALL processes must be able to reserve swap space - not use it - just put a placeholder there in case it needs to use it. You *must* have swap space equal in memory space to all processes currently running or wanting to run.

When you say "equal in memory space", are you referring to physical memory? As I stated in mt original post, swapingo shows !G, but in the system properties in SAM, this shows 8G. Which memory space are you referring to?

Thanks.

[Jeff Schussele]

Hi (again) James,

I mean just exactly how much swap space do you have? This would include primary - usually /dev/vg00/lvol2 - and any/all secondary swap space.
Run the command & it will tell you how much.
Now, if it's low & you cannot add any more, then you could set swapmem_on = 1 to get 75% of RAM as pseudo-swap to help you get more reservation space.

HTH,
Jeff

[John Palmer]

The Oracle alert log should give some indication about why your instance won't start. Please post any error messages that it contains.

Regards,
John

[Jeff Schussele]

I'm sorry, I didn't directly answer your question.
You must have reservation swap space >= the sum of the memory requirements of *all* the processes that are running or want to run.

Rgds,
Jeff