Using Oracle 9 . Why the system will be paging out when there is still
plenty of free memory about 7GB?

Box is running 3 instances of oracle.
last pid: 25158; load averages: 4.72, 5.64, 6.30 12:28:49
477 processes: 467 sleeping, 5 zombie, 5 on cpu
CPU states: 80.0% idle, 8.5% user, 3.9% kernel, 7.6% iowait, 0.0% swap
Memory: 14G real, 7447M free, 6834M swap in use, 5846M swap free

Why would swap being used instead of real memory? Could that have been
instructed to used swap ? How to know. Thanks.

Here is buffer cache, is it too large? how about Shared memory?
sysdef info
292241408 maximum memory allowed in buffer cache (bufhwm)
4294967295 max shared memory segment size (SHMMAX)


vmstat  3|awk '$9>79 || $9 == po'
 procs     memory            page            disk          faults
cpu
 r b w   swap  free  re  mf pi po fr de sr m0 m1 m3 m5   in   sy   cs
us sy id
 0 1 0 4745184 7378840 13 1283 32 80 80 0 0 0  0  0  0 3700 26842 3024
14 6 80
 0 2 0 4647192 7343464 637 1738 3485 88 88 0 0 6 0 0 0 3725 156306 4402
28 9 63
 procs     memory            page            disk          faults
cpu
 r b w   swap  free  re  mf pi po fr de sr m0 m1 m3 m5   in   sy   cs
us sy id
 0 5 0 4827904 7406088 501 857 26 82 82 0 0 1  0  0  0 6140 30742 9697
22 9 69
 0 4 0 4861800 7410728 39 91 26 130 130 0 0 0  0  0  0 5199 10508 4670
11 4 85
 0 0 0 4848664 7401880 17 61 0 80 80  0  0  0  0  0  0  879 153171 1533
4 3 94
 procs     memory            page            disk          faults
cpu
 r b w   swap  free  re  mf pi po fr de sr m0 m1 m3 m5   in   sy   cs
us sy id
 0 0 0 4791752 7392448 36 285 0 0  0  0  0  0  0  0  0 1502 173372 1951
4 6 91
 0 6 0 4471272 7319752 86 1133 16 98 98 0 0 6  0  0  0 4780 14291 4987
16 4 80
 0 4 0 4452664 7315928 108 921 32 213 213 0 0 0 0 7  0 4225 19746 3747
16 5 79
 0 4 0 4429464 7313440 216 1818 16 125 125 0 0 3 0 0 0 4284 38156 4356
14 7 79
 0 3 0 4432856 7313720 43 801 10 106 106 0 0 0 0  0  0 5312 29292 4230
13 7 80
 0 4 0 4422616 7303728 38 857 37 82 82 0 0  0  0  0  0 5270 18917 4413
13 5 82
 procs     memory            page            disk          faults
cpu
 r b w   swap  free  re  mf pi po fr de sr m0 m1 m3 m5   in   sy   cs
us sy id
 0 4 0 4372136 7279400 3 139 5 88 88  0  0  3  0  0  0 5880 18986 5868
15 6 80

MikeHT wrote:
> Using Oracle 9 . Why the system will be paging out when there is still
> plenty of free memory about 7GB?
> 
> Box is running 3 instances of oracle.
> last pid: 25158; load averages: 4.72, 5.64, 6.30 12:28:49
> 477 processes: 467 sleeping, 5 zombie, 5 on cpu
> CPU states: 80.0% idle, 8.5% user, 3.9% kernel, 7.6% iowait, 0.0% swap
> Memory: 14G real, 7447M free, 6834M swap in use, 5846M swap free
> 
> Why would swap being used instead of real memory? Could that have been
> instructed to used swap ? How to know. Thanks.

Have a look for the reply by Casper Dik under the recent post "Why using 
swap space instead of real memory?" That explains why top is showing the 
swap data it does.

The fact your scan rate is low indicates you have sufficient memory.

'top' is great, but its data on memory usage is not too useful. I think 
it is true to say that if you are low on memory 'top' will indicate 
that, BUT you can have plenty of RAM and top will still indiate a low 
value for free memory.

I don't know how easy it is to read the scan rate with a C program 
(without having a PhD on Solaris internals), but it would be useful if 
top showed the scan rate as one of its parameters. At least then it 
could become useful for looking at memory usage, but without that, it is 
next to useless for that task.

-- 
Dave K

http://www.southminster-branch-line.org.uk/

Please note my email address changes periodically to avoid spam.
It is always of the form: month-year@domain. Hitting reply will work
for a couple of months only. Later set it manually. The month is
always written in 3 letters (e.g. Jan, not January etc)

Dave <INVALID.See-signature-for-how-to-determine@southminister-branch-line.org.uk> wrote:

> 'top' is great, but its data on memory usage is not too useful. I think 
> it is true to say that if you are low on memory 'top' will indicate 
> that, BUT you can have plenty of RAM and top will still indiate a low 
> value for free memory.

As far as I've seen, top accurately reports the OS's value for
'freemem', but that value has different meanings in different OS's.

Prior to Solaris 8 the filesystem buffer cache directly claimed buffer
memory from the freelist, so any system using I/O always had a low value
for freemem.  For Solaris 8 and up, this is changed and the figure
should be a much better indication of the actual free memory on the
system.  

> I don't know how easy it is to read the scan rate with a C program 
> (without having a PhD on Solaris internals), but it would be useful if 
> top showed the scan rate as one of its parameters. At least then it 
> could become useful for looking at memory usage, but without that, it is 
> next to useless for that task.

Scan rate should fall into the same category.  Standard I/O will drive
the scan rate up in pre-8 Solaris even if there's plenty of memory, but
it shouldn't in 8 and higher.

I don't see why it's useless.  I've got a machine here where both 'top'
and 'vmstat' report about 4GB of RAM free.  Seems about right to me...

-- 
Darren Dunham                                           ddunham@taos.com
Senior Technical Consultant         TAOS            http://www.taos.com/
Got some Dr Pepper?                           San Francisco, CA bay area
         < This line left intentionally blank to confuse you. >