Performance degradation in Admission control

[baber]

Hi

can you explain more clearly about Performance degradation Warning message in Admission control:

Because while i set HA cluster on percentage or Host failure cluster tolerance  on 1 host it means my HA will be work and that's tolerate is 1 Host now why we have to set Performance degradation ?

would you please give a example

BR

[sk84]

The ESXi hosts within a cluster do not necessarily have to be identical. For example, one host can have 128 GB RAM, another host only 64 GB RAM. If the larger host fails, bottlenecks could occur on the smaller host.

Another example is the overcommitment of resources. If one host fails, more VMs run on the other hosts, so that the overcommitment (especially for the CPU resources) becomes higher. See vCPU : pCore ratio.

[baber]

Thanks

but could not understand what is the benefit of

Performance degradation in Admission control

would you please take an example for example what is difference if i set that on 30% or 50%

BR

Babak

[sk84]

If you specify that the cluster can tolerate a host failure, it means that a certain percentage of all cluster resources is reserved to cope with this failure. For example, if you have 3 hosts in a cluster and 1 host is allowed to fail, 33% failover capacity will be kept free on all hosts. This means that each host can only be used to 67% of its capacity.

However, only the reserved memory and CPU are included in this failover capacity calculation, and a virtual machine can use more memory than is reserved (for example, if only 50% of the allocated memory is reserved).

The performance degradation setting takes this into account and is intended as a threshold value.

For example:

All VMs have reserved 50% of the memory and/or CPU resources, but use twice as many resources. The hosts are used as much as possible. If you have now set the performance degradation to below 50%, an alarm will appear due to an HA configuration error and insufficient resources. The VMs will be started, but will only get 50% of the resources (that's what was reserved) and not what they originally had in use. The infrastructure will inform you about this, as it will cause most likely performance issues.

If you have set this value to over 50%, no alarm or warning will appear as you indicate that these resources bottlenecks are okay for you in case of a host failure.

[diegodco31]

Hi

See this link

vSphere 6.5 what’s new - HA - Yellow Bricks

[baber]

thanks

but still confuse.

imagine i have 20 host in my cluster and i set Host failures cluster tolerates = 2

now what happen will be do i f i set Performance degradation VMs tolerate 10% , 20% , 50% , 100%

BR

[ThompsG]

Hi baber​,

So as you know this is about "Admission Control". The "Performance degradation VMs tolerate" setting allows vSphere to determine that if powering on a new VM would mean that in the failover event (host fails) the other VMs would suffer a degradation in performance. In this case a warning is raised to tell you that VM performance might suffer in a failover event.

The slider simply allows you to determine how agressive before you get a warning with 0% meaning a failover event will cause no performance issues and 100% meaning the warning is diabled.

While these setting could seem redundant, think about it from the prospective of if you don't have reservations configured. Without reservations you could deploy VMs without realising in a failover event things may not run as you expect. With this option enabled you at least get a heads up.

Does this make sense?

[baber]

As I understood, supposing I have 20 esxi hosts and set Host failures cluster tolerates  =2,  now 2 hosts will be down and i set Performance degradation VMs tolerate =0% all of the vm will be restart on the other hosts and now in this mode if i create a new vm (while 2 hosts are failed) if i don't have sufficient resource it will be show a warning or alarm whereas will create new vm and power.on with low performance.

Is that correct ?       

[ThompsG]

Yes and no

Lets step through it (just based on memory):

1. All 20 of your hosts running happily and everybody is having a great day (Lets say you have 200GB of available physical memory across the hosts)
2. Host failures to tolerate are 2 (cluster sets aside/reserves 20GB of memory to allow for guests restarts when hosts fail) NOTE: Active memory is not reserved memory
3. Memory actively used is 190GB across the cluster
4. Percentage of performance degradation is set to 0 (zero) %
5. You go to deploy a new VM to the cluster
6. vSphere Availability checks to see what is currently being consumed within the cluster against what resources the cluster would have if 2 hosts failed (190GB currently needed but would only have 180GB available IF 2 hosts were to fail)

In the above case you get a warning as there would likely be a performance issue in a HA event.

Does that help?

[baber]

As I understood i have 20 hosts that each of them has 10G ram eventually i have 200GB memory now two of my hosts fail and in this step i have 180G memory now in this mode if i create a vm

with 10GB memory it will show a warning

is that correct ?

BR