Not enough space for Oracle shared memory segment

[ALH]

During the shutdown procedure of our Oracle database we get the message:
SVRMGR> ORA-07306: sms1sg: shmget error, unable to get a shared memory segment.
HP-UX Error: 12: Not enough space

Our shutdown procedure is:
connect internal
shutdown abort
startup restrict
shutdown immediate

With the startup command we get the error.

My question is now:
What is the reason of this errormesage and what for relation is there between Oracle en hp-ux for deleting the shared memory segment.
Is there a fix if this problem is structurel.

we have hp-ux 10.20 en Oracle 7.3.4.4.1

Thank You
Cees Wielink

[Rob Smith]

Hi, I am not a DBA but there may be something wrong with the way you are shutting down. Obviously it is not releasing it's shared memory segment. I have had this problem on a couple of occassions running Oracle 7.3.2 and the only way I could get it back up was to reboot the server. Hope this helps.

Rob

[ALH]

Additional info:

After this error the database was backed up. After the backup there was a script started that starts up the Oracle database. This was done without any errors. Only a crash recovery of oracle. The database runs OK.
So, was the delete of the shared memory segment slow??????

[Volker Borowski]

Hi Cees,

check what else is running on this box, you might need to extend swap and kernel parameters.

Volker

[Mladen Despic]

My 2 cents:

When Oracle is shut down, other processes running on the system may be locking memory thus not leaving enough for Oracle to restart properly. A system reboot will likely fix this, but you will probably want to avoid rebooting the system every time you want to restart Oracle.

Is your pseudo-swap enabled? (If 'swapinfo -tam' displays a "memory" line, then your pseudo-swap is enabled.) If it is, it may be competing with Oracle for locked memory. In any case, Oracle recommends that pseudo-swap be disabled on a system running Oracle database, i.e., change the value of the kernel parameter swapmem_on from 1 to 0. Before making this change, make sure that there is enough swap on disk.

[Kieran Lismore]

Hello!

My suggestion is to verify the max number of semaphores on your kernel and also the limit of max shared memory segm.

You can check the shared memory/semaphores using the command "ipcs -a".

[Wodisch]

Hello ALH,

a couple of things to mention:
1) you should insert "alter system checkpoint"
into your commands to speed up the startup:
connect internal
alter system checkpoint;
shutdown abort
startup restrict
shutdown immediate
2) it seems to be an old and re-occuring error
that semaphores are not released on shutdown
check by including "!ipcs -sb -mb" after the
"shutdown abort" command in your list. If
you see something owned by "oracle" (or
whateve is the name of your local "owner"),
especially when there is a "D" as first
character of the "permissions" (=Delete),
then that entry is deleted - but still there
3) you might really running out of free RAM or
swap, so test check wether your instance is
using the parameters "lock_sga=true" and/or
"pre_page_sga=true". and insert a comand
like "!vmstat 1 5" or so into your command
chain

HTH,
Wodisch

[Magdi KAMAL]

Hi

Beside you are using "Shudown abort" which is really dangerous, you should increase your swap space to 3 or 4 times the amount of your memory.

What could happend that, while the hp-ux system is still releasing the shared memory segments due to your "shutdown abort", it gets a new request for "startup restrict" which asks for the same amount of memory segments which hp-ux could not get from memory and even could not reserve space in the swap space.

for me 2 points :

1. Change you strategy of "shutdown abort" and "startup restrict".
2. Control your swap size.