Instance Caging

Last week we had a funny issue at a customers ODA X7-2M. You may know, that there are different database shapes, each having different settings for CPU_COUNT, memory etc. The documentation is quite clear about that. In this special case they had a database with shape “odb4” which sets CPU_COUNT to 4. Now they went live with their appication on that database and experienced the following behaviour.
cpu_count_instance_caging
You see the CPU limit of and you can also see, that the load goes well over that limit. But why? The answer is quite simple. Setting CPU_COUNT does not neccessarily limit the number of CPUs the database will use. An active Resource Manager Plan is required to achieve that. MOS Note “Why are a Higher Number of CPUs Used Than the CPU_COUNT Parameter Setting (Doc ID 2017395.1)” is explainig that. If you wish to limit the CPU usage, you will have to use Instance Caging which is a combination of setting CPU_COUNT and using Resource Manager. There is another MOS Note “Configuring and Monitoring Instance Caging (Doc ID 1362445.1)” for that.
That means, if you want your databases on ODA be limited to the CPU usage their shape defines, you need to activate a Resource Manager Plan, at least the “DEFAULT_PLAN”. This is not the default, keep that in mind.

Advertisements