"IMP" PROCESS BLOCK ON BY "IPC"?

[huaxiang]

attache show "imp" process block on by "ipc", import data into database seems slow, what does that mean? shoud any kernel be adjusted?

[Duncan Edmonstone]

Well I'm assuming this is an 11iv3 system as vhand has some activity depsite there being no memory pressure.

There doesn't appear to be any resource issues at all on this system (CPU/memory/IO). The only thing you don't see here is if there is pressure on some of the IPC resources (given your wait states perhaps you should check that)

Pressing 't' in glance will show you if you have contention there.

That said, more likely these processes are just waiting on a semaphore before continuuing work, which would point to something in Oracle. So unless any of your system tables are showing pressure I think this is one to look at from the Oracle perspective.

One other point - glance seems to be taking an awful lot of CPU time (although you have plenty to spare). I'd look at applying the latest patch bundle (Sep 08) to this system as well.

HTH

Duncan

[Steven E. Protter]

Shalom,

Each oracle process appears to be using 103 MB of ram.

It will not take long for you to run out of memory under these circumstances.

Check kernel params.

shmmax

Make sure it meets guidelines on your version of oracle.

Make sure it is adequate.

check shared memory processes. If any oracle shared memory areas are owned by root its game over, reboot time, oracle can not be run by root.

Also, your post.

No OS version
No Oracle Version

Lots of information that that could get you better help, missing in action.

SEP

[TTr]

I think the question here is what does "blocked on IPC mean"? IPC refers to the interprocess communication segments which are shared memory, semaphores and messages. The import process is blocked because it needs to wait for an ipc segment (shared memory or semaphore) to become available so that imp can use it before it can continue. There are probably important oracle jobs running that are keeping the database busy Check your database to see what kind of activity is taking place and find out what you can do to make the import process continue. Also check what the imp processes are doing, they may be trying to enter or overwrite data that is active and they can not. They may never be abble to start. Ensure that the imp jobs are doing something that is correct.
Also why do you have that many imp processes? Sometimes these imports will become serialized because of their own nature, (depending on what and where they import in the database), that is they have to run in sequence, when one of them completes a second one can start.
I think this is a database issue more than a kernel tuning issue.

[huaxiang]

thank u all. we are now checking "imp" process;
os version: 11.31,oracle: 10g;