VMware Horizon Community
chriskoch99
Enthusiast
Enthusiast

How does CPA pick which local pool to use for a global entitlement?

I know there are the three options -- All Sites, Within Site, Within Pod -- and I get what each of those mean.   But, for example, if I choose Within Site, and I've got three pods in that site, under what circumstances does a broker send the user to a VDI outside of its local pod?

For example:  Horizon 7.13 on-premises.  In Site A, I've got Pod1, Pod2, and Pod3.  Global Entitlement is set to "Within Site".  Brokers from Pod1, Pod2, and Pod3 are in a hardware load balancer VIP that users hit.

So if a user gets sent to a broker in Pod1, it looks like that broker will have a tendency to send them to a VDI in Pod1, but exactly when will it send them to another pod?  Is it only when the local pool in Pod1 is not available?   Is there some capacity threshold over which the broker will know to send users to a Pod2/3 pool in the same GE?   Is the hardware load balancer the only way to ensure even session distribution over the three pods, or are the brokers intelligently handling some of this as well?

Thanks!

Reply
0 Kudos
5 Replies
BenTrojahn
Enthusiast
Enthusiast

when the local site is not available, OR doesn't answer.  AFAIK its still not like citrix that will "try again".  

Try the local pools "x" time is a feature i have asked for, but im not sure if it made it to 7.13 or later.  

Reply
0 Kudos
chriskoch99
Enthusiast
Enthusiast

So, it doesn't seem to be bearing out exactly this way for us.  

 

We have a Global Entitlement with desktop pools in Pod1 and Pod2.  We have assigned around 35 users to the GE, but they are only connecting to brokers on Pod1.  Pod2 brokers are not yet part of the load balance.   

 

If what you're saying is true, we'd expect all users to get sent to Pod1 VDIs.  But in fact, we see three users have been sent to VDIs in Pod2 and everyone else sent to Pod1.   Which doesn't make sense given there have been no availability or communication issues with Pod1 since we added everyone to the GE.

 

So it seems like some other logic may be in play here?  Why would those three users have been sent to Pod2?

Reply
0 Kudos
sjesse
Leadership
Leadership

Do they have a  homesite

Using Home Sites (vmware.com)

Also if there are issues getting into the pod they are assigned they will be redirected to the other pod.

 

 

Reply
0 Kudos
chriskoch99
Enthusiast
Enthusiast

No home sites assigned, because today we're using CPA with multiple pods in a single site/datacenter.  So there's only one site in play.  Later, when we stand up a pod in AVS, we'll add a DR site and deal with home sites.

Reply
0 Kudos
BenTrojahn
Enthusiast
Enthusiast

From my reading of the server logs it will try the pools in the active pod *once* (the one your DNS lb vip is pointing to) then it will try the other pools in CPA.  Since both pods have pools that are part of the Global you will have the chance to launch from both pods *occasionally* as i described.   Want to give your users another chance to launch in your preferred Pod?  Add another pool in that pod to the global or split your single pool up.

 

Reply
0 Kudos