Strange memory errors

[Bruno Bossier_1]

Our OVO console lately show us the following error :

Monitor _GlbSwapBytes: Threshold: 50.00 Value: 87.76

The Memory resource is under stress, processes are swapped out/deactivated.

11/23/05 08:59:33 severity warning

Monitor _GlbPageScanned: Threshold: 200.00
Value: 557.25

The Memory scan rate is quite high, currently
557.25.


I find this strange, because this server is a rp4440 with 6GB physical memory and 4GB additional swapspace. It is running 2 Oracle 10g instances. This node is also part of a Serviceguard cluster.

I'm hesitating to add another 4GB of swapspace as I'm not really sure of this will solve the problem.

Any ideas ?

[Mark Nieuwboer]

Bruno,

Did you check with top our glance wich process takes the memory. Maybe only 1 process is going wild.

grtz. Mark

[Bruno Bossier_1]

It's not quite clear. Here is also the output of swapinfo :

Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4194304 1168776 3025528 28% 0 - 1 /dev/vg00/lvol2
reserve - 3025528 -3025528
memory 6391552 5417636 973916 85%

[Mark Nieuwboer]

You don't need to extend the swap it's quit big and i know oracle want 2x the size swapspace in relation with the memory.
But i think it's more a oracle setting problem then a memory problem.

grtz. Mark

[Bruno Bossier_1]

I talked in the meantime with our Oracle dba'er and told me that the memory size that is configured in Oracle is too high. He is going to change that and stop/restart the database.

I will see whether that helped.

Anyway, thanks for the help !

Cheers,
Bruno

[Chuck Dennis_1]

Your swap space is under configured since you have a system with 6GB mem, but only 4GB device swap allocated. Resizing Oracle will help with the memory pressure, but you should increase swap to a minimum of 1X memory...and preferably to 2X (or more).

[Steve Steel]

Hi

ftp://eh:spear9@hprc.external.hp.com/memory.htm


memory tool heaven with a good explanation


Steve Steel

We make house calls and bring our own bugs.

[Ted Buis]

Look for page outs in glance. Did you change the default size of the dynamic buffer cache. If dbc_max_pct is still at 50%, then it is way too high. If you have OnLineJFS then you can set up mount options to avoid double buffering by the OS and Oracle. See the attached cookbook.