Host memory usage Warning

[MariusRoma]

An ESXi 5.x host member of a vSphere 5.x infrastructure and used in a lab area to host VMs used to test applications reports an Host memory usage Warning.

Memory usage is about 118 GB with 131 GB physical RAM.

Here are some detailed values:

• Baloon valuee is 0
  
• Granted Memory is about 119 GB
  
• Consumed memory is about 120 GB
  
• Active Memory is about 10 GB
  
• Swap used is 0
  
• Shared common is 3 MB
  
• Compressed is 0
  
• Shared is about 1,3 GB
  
• Overhead ios 1,8 GB
  
• Total Capacity is 125 MB
  

Based on what I learned, value 0 for Baloon sounds good, and I should not worry about memory use...

On the other hand I would expect a greater value for Shared as many VMs are very similar and use exactly the same OS version.

Looking at Resource Allocation for a typical guest VM I see:

• Host memory: 6 GB consumed, 60 MB overhead
  
• Guest memory: 6 GB Private, o Balooned, 0 shared, 0 Compressed, 245 MB active.
  

Should I worry about the Host Memory usage warning?

Should I suspect any configuration error either at host and/or VM level?

Is there any document explaining how to troubleshoot memory issues?

Regards

marius

[aaronwsmith]

Your shared memory value looks pretty good to me.  I personally don't see shared memory that high on our hosts.

Whether or not you should worry about your host memory usage depends on your setup.  Is the host part of a cluster?  If so, how are you planning to handle capacity/load if a host fails (i.e. N+1 architecture)?

You might want to check your VMs to see if any are grossly over allocated on memory.  Active memory counter is an estimated metric by the vmkernel as it watches which memory pages the guest OS is accessing.  But it may not reflect how much memory the VM really needs.

You can also set limits on some of those VMs to force some of their memory to vswap, if you're okay with the performance trade-off.  That way the VMs are still presented the same amount of RAM, but a certain (hopefully unused) portion is swapped to physical disk by the hypervisor.

Also - Check the threshold for the vCenter alarm triggering the memory usage warning and see if it's set more conservative than you're worried about.

Hope this helps!

[lenzker]

"You can also set limits on some of those VMs to force some of their memory to vswap, if you're okay with the performance trade-off.  That way the VMs are still presented the same amount of RAM, but a certain (hopefully unused) portion is swapped to physical disk by the hypervisor."

I wouldn't recommend this. Don't use Limits except you really have a good use case for it ( I can't imagine one for memory ). Swapping is bad for your VM and might create a lot of IO stress on your storage. Use shares instead if you want to prefere some of virtual machines.

The amount of shared pages really depends on your OS, applications and if your are using large page tables.

Do you have this low active value all the time or just sometimes? If it's that low all the time you probably should think about resizing your VMs and put less memory in it. This has to planned very carefully.

I don't know about any document talking about troubleshooting memory issues (you can check this general doc here: http://www.vmware.com/pdf/Perf_Best_Practices_vSphere5.1.pdf ), but usually there are two solutions.

"less v(irutal)Ram, or more p(hyiscal)Ram"