RE: reading the SGA from my own program
- From: John D Parker <orclwzrd@xxxxxxxxx>
- To: jeremiah@xxxxxxxxxxx, 'Anjo Kolk' <anjo.kolk@xxxxxxxxxxx>
- Date: Fri, 18 Aug 2006 19:11:32 -0700 (PDT)
Jerimiah, have you read Kyle's chapter in the oak table book? It's all in
there. You can't just find the address of one end and step off the distance.
The individual session addresses are all in the array that you referenced
first. Step thru that array and read the addresses listed there with the proper
structure and it works. I've written a simple sga program and it works well.
Much to my surprise, sessions were used from the high end to the low end. I set
limits in my test program and kept getting zeros back. I took the limits out
and discovered that the high order sessions were used first much to my
surprise. Unfortunately at that point real work intervened and I haven't had a
chance to get back to it...
John Parker
Jeremiah Wilton <jeremiah@xxxxxxxxxxx> wrote: Anjo,
I guess I realize I could set shmmax high and get a single segment, but I
want my program to work on any old Oracle instance, not just one I set up
special for it.
Are you saying that because segments can be noncontiguous, you have to count
the space between segments where there is no segment allocated when you are
counting up from the SGA base address to the address of the object you are
seeking?
In my case the segments are noncontiguous. There are big gaps between each
segment according to oradebug IPC. If I count space between segments, then
the supposed location of session waits falls in the last segment. I'll play
around and see what I find. Any hints are welcome.
Is there a better (more programmatic way) to map the segments' base
addresses and sizes other than looking at the trace file from oradebug ipc?
Thanks,
Jeremiah Wilton
ORA-600 Consulting
http://www.ora-600.net
________________________________________
From: akolk@xxxxxxxxxxx [mailto:akolk@xxxxxxxxxxx] On Behalf Of Anjo Kolk
Your assumption is wrong. The fixed SGA contains variables that are known at
oracle compile time or point to structures in variable part. So the answer
is that generally speaking, all x$ are in the variable part of the SGA. Also
there may be multiple segments but they are contiguous in memory (not always
the case), so you will have to map/attach at least to the segments of the
fixed and variable part. A better trick you can pull is so set SHMMAX
extremely large, so the whole SGA will fit in one segment. That means that
every thing is attached (easy to do when you are on linux).
Anjo.
--
//www.freelists.org/webpage/oracle-l
---------------------------------
Do you Yahoo!?
Everyone is raving about the all-new Yahoo! Mail Beta.
Other related posts:
