I frequently receive alerts via EM 12c that I have reached near 100% of my memory for my host servers. My servers have lots of RAM and have been running fine for a long time. I have tried to increase the metrics so it does not warn me until it hits 99.5% but it still sends me alerts. I am not sure what I have to configure differently for EM to not think this is a problem.
ok. That was worth checking I will email the host team offline to look into this thread.
BTW - are you saying that the data is not reporting high values but you are getting alerts or is it that the data is reporting high values and you want to understand where the data points are coming from?
If you could upload a snapshot of the metric from the all metrics page - it might be helpful to the host team.
Just to clarify a few things. My oracle databases instances are running on an 2 HPUX server. Enterprise Manger consistently shows that the Memory uitilization is near 100%, it never really drops it just stays flat lined at near 100% and every once and a while their will be a spike where it will drop to 90 or 95%. I am letting AMM auto tune the memory between PGA and SGA. I only have agents running on HPUX for Enterprise Manager.
I have Enterprise Manager 18.104.22.168 setup on a Windows 2008 R2 server. I have only had EM setup for a few months but I beleive the issue of memory showing near 100% is an old issue. Again the environment and the servers are working fine.
After doing a bit more research to me the real problem is how enterprise manager looks at the memory. If I use an utility such as a tool that comes with HPUX called "Glance" it shows the following. I have 32 GB or Physical memory, System is using 7.5 GB of this memory and User is using 10.7GB and their is 15.1 GB of free memory. So the "Glance" utility shows my memory utilization is just fine.
FYI memory is also showed to be healthy when looking at it with the below command.
# swapinfo -m
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 34816 0 34816 0% 0 - 1 /dev/vg00/lvol2
dev 34816 35 34781 0% 0 - 0 /dev/vg02/lvolsw1
reserve - 6626 -6626
memory 31147 14588 16559 47%
To me it seems to me that Enterprise Manager is adding up the System Memory + User Memory + Free Memory and comparing it to the Physical Memory. When it should be taking only System Memory + User Memory and comparing it to Physcial memory. So the metrics it is looking at need to be modified somehow.
It is quite normal for memory to be at 100% from an OS point of view. By default oracle occupies that memory even if it is not actually using it.
in say oracle 8 we used to manually set how much memory is used for sga, java etc etc and as a rule of thumb we used to give oracle about 2/3 of system memory and leave 1/3 for os and other applications.
Now everything is automatically taken care of, including memory allocations, and it is obvious that oracle will try to make use of every available resource, in order to deliver the best performance.