Apologies if this is a repeated question but i am unable to find any answers any where.
I am facing a typical situation wrt to capacity reports getting generated in VROPS 6.4 .
While the reports are showing 0 % available resources for one particular cluster, There is at least 30 % memory resources available when i look at Cluster in Vcenter. I realize metrics for memory are not defined correctly but unable to find where to find the metrics definitions . looking into policy, There is 25% over commitment defined for cluster .
How do i correct this so that VROPS send out correct capacity reports showing correct capacity, as available in V center.
Bigger Q - where are these metrics defined based on which capacity is calculated. This is chronic issue in my environment . I tried going through vmware without much results
Cheers
Just to quickly clarify
Vrops can use 3 different capacity models
Going by what you have selected in you policy you are using an allocation based model. This model has some advantages and disadvantages
Advantages
You can define an over commit ratio (the only model you can)
You can define server profiles sized and see how may deployments are left at that size
Disadvantages
Completely ignores what the VMs are using on the hosts ( that is what the other two models will look at but you cant do server profiles)
Now lets look at vCenter
vCenter reports usage based on Consumed not allocation. You can have a VM with 16 GB ram and in vrops it will report that that VM has used 16GB ram (allocation model) but in vcenter it could only show the VM using 2GB of ram because the VM has not accessed the remainder of the memory in a wile so ESXi has reclaimed it to be used by other VMs (standard ESXi memory management). If you think of vcenter reporting on "Hot" memory that is dynamic and vrops allocation model capacity is using a hard allocation limit with hard set ratios that never change regardless of what the VM is consuming or demanding.
So i would say vrops is reporting back correctly based on you policy but you cannot compare the results againest vcenter because it is reporting back usage in a completely differently.
The only way to get anything close to what vcenter is displaying you will need to change your capacity model to consumed ( you can enable all 3 and transition to the desired model). I personally like the allocation model as i will always leave enough capacity to increase the VMs after the cluster has become full based on the allocations left with out having to move them.
Sorry i dont understand the Bigger q
Just to quickly clarify
Vrops can use 3 different capacity models
Going by what you have selected in you policy you are using an allocation based model. This model has some advantages and disadvantages
Advantages
You can define an over commit ratio (the only model you can)
You can define server profiles sized and see how may deployments are left at that size
Disadvantages
Completely ignores what the VMs are using on the hosts ( that is what the other two models will look at but you cant do server profiles)
Now lets look at vCenter
vCenter reports usage based on Consumed not allocation. You can have a VM with 16 GB ram and in vrops it will report that that VM has used 16GB ram (allocation model) but in vcenter it could only show the VM using 2GB of ram because the VM has not accessed the remainder of the memory in a wile so ESXi has reclaimed it to be used by other VMs (standard ESXi memory management). If you think of vcenter reporting on "Hot" memory that is dynamic and vrops allocation model capacity is using a hard allocation limit with hard set ratios that never change regardless of what the VM is consuming or demanding.
So i would say vrops is reporting back correctly based on you policy but you cannot compare the results againest vcenter because it is reporting back usage in a completely differently.
The only way to get anything close to what vcenter is displaying you will need to change your capacity model to consumed ( you can enable all 3 and transition to the desired model). I personally like the allocation model as i will always leave enough capacity to increase the VMs after the cluster has become full based on the allocations left with out having to move them.
Sorry i dont understand the Bigger q
Hi,
Thanks for revert . It is true that we are using allocation based model for capacity calculations . Making some sense to me now as i was thinking that VROPS is calculating capacity based on active VM memory rather than consumed memory.
Given the fact that we want VROPS to show us as exact status as it is in VC , Which is the preferred model for capacity calculation. Based on your experience i believe "memory Consumed" will give us a better "on the ground" picture or is it possible to choose both allocation models for capacity calculations and future capacity projections .
Hoping that having the both allocation methods will give us a more realistic picture of capacity .
As for Bigger Q goes, I was interested in seeing where the memory metrics are defined. Perhaps some where in Admin console .
Cheers
You are defining the metric to use when you configure the policy. There are Total capacity, Usable capacity and remaining capacity metric for all three model in vrops
allocation based memory for example
Total capacity = Sum all the memory in the hosts and * by you over allocation ratio
Usable Capacity = the above - any buffers you set up
Remaining capacity = Usable capacity - all the memory allocated to all the VMs in the cluster
Use the same for CPU
If you look under the memory metric against any capacity container ( Host, cluster, CDC etc) you will see 3 sub folders one for allocation one for demand and one for consumed all have there specific metric
I dont use the other two models so not 100% what is used for them.