OCI8 drivers causing a regular core dump on HP-UX 11.11 PA-RISC

[Gousset]

We have a 20 CPU, 20 GB Memory HP-UX PA-RISC server which is core dumping very often with our application that connects to a oracle 9.2.0.7 database via OCI client. We will be moving to thin drivers, but it appears that the Java heap is allocated from the global address space as does the stack space and space for native libraries. The error is pointing that there was no address space from which the memory for native libraries would be allocated. Can someone tell me how to check this global address space and how to increase this value.

[Steven E. Protter]

Shalom,

Suggestions:
Make sure all required patches are in and the system is in general well patched.

Make sure the required libraries are up to date.

Global address space is addressed by kernel parameters. shmmax shmmni, etc.

SEP

[Gousset]

Hi,

Our current kernel O/S is as follows:

shmem = 1
shmmax = 5637144576
shmmni = 512
shmseg = 120

Do you think we would need to increase any of the above ?

[Dennis Handly]

>The error is pointing that there was no address space from which the memory for native libraries would be allocated.

Where does it say that?

The siginfo and register dump and instructions indicate that you are calling $$dyncall_external and passing a null plabel (R22). That references si_addr=00000002 on
>p 0x0ec41093\i
LDWS 2(r22),r19

Somewhere in Oracle: libclntsh.sl.9.0+0x1e15fc

[Gousset]

"The error is pointing that there was no address space from which the memory for native libraries would be allocated"

Actually refers to the error message we recieved on our application when we tried to use 9208 client jdbc/oci libaries. The problem seems to be related to oci drivers. Can you conclude anything by looking at the hs_err_pid file ?

[Dennis Handly]

>Gousset: Can you conclude anything by looking at the hs_err_pid file?

Probably not, I have no ideas about that file. I just know the low level reason you got that signal 10.