Shared memory segments on HP-UX 11.23/9i

[Brendan Newport_2]

Hi!

I have two nodes in a RAC/ServiceGuard cluster that have the SGA extended from 6GB to 25GB.

One of the nodes is displaying just one 25GB memory segment (good);

m 7 0x28e74644 --rw-r----- oracle dba oracle dba 33 26863513600 2409 29506 11:08:30 11:10:31 13:50:

But the other nodes is displaying 5 segments, once again totalling 25GB, but not all equal sizes;

m 10977285 0x00000000 --rw-r----- oracle dba oracle dba 32 13438550016 18686 28497 11:14:57 11:16:04 11:46:32
m 950278 0x00000000 --rw-r----- oracle dba oracle dba 32 6727663616 18686 28497 11:14:57 11:16:04 11:46:32
m 163847 0x00000000 --rw-r----- oracle dba oracle dba 32 3372220416 18686 28497 11:14:57 11:16:04 11:46:32
m 163848 0x00000000 --rw-r----- oracle dba oracle dba 32 1694498816 18686 28497 11:14:57 11:16:04 11:46:32
m 9 0x1da2dd34 --rw-r----- oracle dba oracle dba 32 1697689600 18686 28497 11:14:57 11:16:04 11:46:32

The kernel tunables (such as shmmax and shmmni) are identical on both nodes, but I can't distinquish why one node is using a single 25GB segment as anticipated, and the other five.

Any assistance would be gratefully received.

Bren

[Steven E. Protter]

Shalom Bren,

The difference is due to fragmentation.

When the machine with the broken up shared memory segments went to load oracle, there were other products in shared memory, preventing a contiguous allocation.

Many products, even OS software need and use shared memory.

See whats installed and running on the two systems for further analysis.

SEP