memory lock

[clement_10]

Hello,

I have system running Oracle 8.1.7 on HPUX 11i which does dealocate the memory properply.The system does not release memory allocated to the database instance even after the databses have been stopped.when i do a top i get very low free memory available. I have attachd a copy of top and swapinfo O/P for further reference. Could anyone out there with any usefull information about this problem assist.


Clement O

[T G Manikandan]

In the top output you can notice lot of java processes taking up too much memory.

You can do a

UNIX95= ps -e -o ruser,pid,vsz=Kbytes|sort -nrk2

to check the each process memory usage and find out the one using the memory.


Revert

[clement_10]

the result i have got from UNIX95= ps -e -o ruser,pid,vsz=Kbytes|sort -nrk2 indicates that memory is mostly used by processes owned by oracle users.
Are there some tunnings to be done on Oracle?

[clement_10]

the result i have got from UNIX95= ps -e -o ruser,pid,vsz=Kbytes|sort -nrk2 indicates that memory is mostly used by processes owned by oracle users.
Are there some tunning to be done on Oracle?

[James Murtagh]

Hi Clement,

If I understand this right you are saying Oracle is not freeing its shared memory after the DB has been shut down?

If so you could use the ipcs command to find the shared memory id in question and the HP contributed too "shminfo" to find which processes are holding the segment. Shminfo can be found here:

ftp://contrib:9unsupp8@hprc.external.hp.com/sysadmin/programs/shminfo

I'll show an example of a shmid waiting for deletion and how shminfo is used to check which processes are holding it.

# ipcs -mob|grep D|tail -1
m 209 0x00000000 D-rw------- www other 6 184324
# shminfo -s 209|grep pid
proc=0x429ce040 (pid 1740 "httpd"): vas=0x42bc4480, SHMEM preg=0x42c2ee00
proc=0x42bc7040 (pid 1739 "httpd"): vas=0x429d7040, SHMEM preg=0x42c0de00
proc=0x42bdd040 (pid 1737 "httpd"): vas=0x42b31940, SHMEM preg=0x42bf4180
proc=0x42a8f040 (pid 1712 "httpd"): vas=0x42b31a80, SHMEM preg=0x42bbe940
proc=0x42957040 (pid 1711 "httpd"): vas=0x42b31800, SHMEM preg=0x42a8b940
proc=0x42b2f040 (pid 1692 "httpd"): vas=0x42b316c0, SHMEM preg=0x42a772c0

You should be able to find the problem from here.

Just a word of note: if the shared memory segment is very large it can take a long time for the memory to be freed and the cache to be flushed. If you are checking memory straight after shutting the database down this may be the cause.

Regards,

James.

[clement_10]

the information i get after running ipcs and shminfo indicate that the httpd deamon does not dealocate the shared memory.During installation of Oracle, the HP version of Apache was disabled and the Oracle version used instead. i tend to think that the Oracle version of apache is causing this memory lock.

[milaren]

i can't find the comm shminfo in my system,why?thx!

[Patrick Wallek]

Milaren,

shminfo is not a standard HP-UX command.

Read through James' reply again and go to the ftp site he mentions to get shminfo and install it.