HA Failover Capacity Levels!

[Dhakshinamoorth]

Hi All,

Please review my cluster status,

VMware vSphere Version=5.0

No.of Nodes in the cluster=4

Each Node configuration=24GB RAM, 12*2.40GHz CPU

Total Cluster Memory=95.96GB

Total Cluster CPU=115 GHz

Configured Admission control policy= CPU:25% and Memory:25%

Cluster have 16 VMs no VMs have CPU reservation and only one VM have 4 GB memory reservation. please find memory overhead for all VMs,

56.91MB,48.81MB,60.84MB,60.84MB,100.22MB,95.28MB,56.91MB,80.07MB,56.91MB,87.33MB,73.43MB,52.89MB,52.95MB,111.52MB,87.33MB,77.02MB.

Please find the attachment and how the CPU and memory failover capacity calculated in percentage?  i reserved 25% means 1/4 node resources available for failover? Hereafter how much resources can i take for new VM installation? anybody reply with calculation? it will helpful to understand more easier. Thanks in advance.

[Sreejesh_D]

Duncan's HA Deep DIve will be a good reading on this topic. It clearly explains the slot calculation and admission control.

http://www.yellow-bricks.com/vmware-high-availability-deepdiv/

[depping]

Read this one:

http://www.yellow-bricks.com/2013/03/14/what-is-current-memory-failover-capacity/

It gives a good explanation I think,

[Dhakshinamoorth]

Hi Duncan,

I need some more clarity on this.

Cluster Summary:

VMware vSphere Version=5.0

No.of Nodes in the cluster=4

Cluster type=Balanced

Each Node configuration=24GB RAM, 12*2.40GHz CPU

Total Cluster Memory=95.96GB

Total Cluster CPU=115 GHz

Configured Admission control policy= CPU:25% and Memory:25%

Current CPU Failover Capacity=96%

Current Memory Failover Capacity=93%

Cluster have 16 VMs and no VMs have CPU reservation and only one VM have 4 GB memory reservation.

please find memory overhead for 16 Powered On VMs,

56.91MB,48.81MB,60.84MB,60.84MB,

100.22MB,95.28MB,56.91MB,80.07MB,

56.91MB,87.33MB,73.43MB,52.89MB,

52.95MB,111.52MB,87.33MB,77.02MB.

I reserved 25% means 1/4 node resources available for failover? i.e 24GB RAM, 28.8 GHz CPU Available for failover?

or CPU=96%-25%=71%=115GHz*71%=81.65GHz CPU available for new VM power on? and Memory=93%-25%=68%=95.96GB*68%=65.25GB RAM Available for new VM power ons?

[depping]

Not sure I understand what you are asking?

[Dhakshinamoorth]

Hi Duncan,

I reserved 25% failover capacity that means 1 node's  resources available for failover out of 4 nodes? i.e 24GB RAM, 28.8 GHz CPU Available for failover?

or CPU=96%(Current failover capacity)-25%(Configured failover capacity)=71%=>115GHz(total cluster cpu)*71%=81.65GHz CPU available for new VM power on?

and for Memory=93%(Current failover capacity)-25%(Configured failover capacity)=68%=>95.96GB(total cluster memory)*68%=65.25GB RAM Available for new VM power ons?

[depping]

Dhakshinamoorthy Balasubramanian wrote:

Hi Duncan,

I reserved 25% failover capacity that means 1 node's  resources available for failover out of 4 nodes? i.e 24GB RAM, 28.8 GHz CPU Available for failover?

or CPU=96%(Current failover capacity)-25%(Configured failover capacity)=71%=>115GHz(total cluster cpu)*71%=81.65GHz CPU available for new VM power on?

and for Memory=93%(Current failover capacity)-25%(Configured failover capacity)=68%=>95.96GB(total cluster memory)*68%=65.25GB RAM Available for new VM power ons?

1) yes with 4 hosts, 25% equals 1 host

2) Read my article.... the percentage will go down to 25% and then vCenter will say you are out of resources. So 75% of your available cluster resources can be used for power-ons.

[Dhakshinamoorth]

2) Read my article.... the percentage will go down to 25% and then vCenter will say you are out of resources. So 75% of your available cluster resources can be used for power-ons.

As per your comment i can only use 75% of cluster resources for power on.but i have only 95.96 GB RAM in cluster and my cluster have 16 powered on VMs with 83 GB allocated memory. how it is possible? sorry to trouble you. still i can't clear.

Total cluster memory=95.96 GB

75% of cluster RAM means=71.97GB RAM (we can power on VMs)

Used memory for powered on VMs=83GB

[vGuy]

As I know the calculation is NOT based on "Used Memory" rather Memory Reservation of the VMs and overhead memory.

[depping]

Allocated --> VMware HA Admission Control is about memory reserved + memory overhead. It has nothing to do with memory allocated. My deepdive HA article explains that in-depth.

[Dhakshinamoorth]

Hi Duncan,

Thanks for continued reply.

one morething. I have 96GB (24GB*4) cluster memory and already 83GB allocated for existing VMs and  if i creating new 4VMs with 4 GB RAM without reservation what will happen?

total physical server memory=96GB

allocated for existing VMs=83GB

if future new 4VMs with 4GB=16GB

96GB<99GB means?