vRA deployment type - Inquiry

[MRoushdy]

Hello,

By going though vRA documentation, VMware says that the "minimal installation" (one vRA appliance + one IaaS) is just suitable for POC and testing, and we should go with the enterprise type by deploying multiple vRAs, and sit them behind a load balancer. However, VMware says that a small environment is that one that has 10 thousand VMs (in the same guides), I may have only 350 VMs, so, is it possible that I go with that "minimal deployment" type? and vSphere HA (or FT) will be supporting against failures, in addition to VM backup indeed.

I need to know your opinion please.

Thanks,

[sk84]

There are different ways of looking at things. Hence the contradictory information in the documentation. First, there is a maximum supported configuration. For example, 10,000 VMs for the minimal setup (is it really that high?). But that doesn't mean that everything runs super smooth and well with 10,000 VMs. Only that it is possible and supported.

There are also different requirements and demands with regard to the intended use. A cloud management and automation platform that is available to customers (internal or external) should usually always be available, because a failure that affects all customers is rarely pleasant.

For these reasons, VMware recommends a distributed setup for professional use. With HA there will be a downtime. Usually between 5-10 minutes until everything is up and running again and customers can use the platform again. A distributed setup is also recommended for performance reasons. The number of VMs is irrelevant. It depends on the workflows running simultaneously. You can have 10,000 VMs and 1 customer makes only 1 change a day. Then a minimal setup would probably be sufficient. In contrast, 50 customers can make changes to 50 VMs at the same time and the minimal setup would quickly reach its limits.

From my own experience, I can say that a minimal setup is also sufficient for productive use if the performance requirements are low and only a few workflows are executed simultaneously. But even then, it can take between 2-5 minutes to complete a configuration change and 5-10 minutes to deploy a new VM.

[MRoushdy]

Thanks for replying. Some where in a guide it mentioned how VMware classifies businesses sizes, 20K VMs is medium, anyway, I forgot to say that I am just a customer company, not a service provider. vRA will be helping us to offer self-service portal to our internal software developers. I just want it to be as simple as possible. A distributed enterprise vRA deployment is not that easy to manage and troubleshoot.

Thanks anyway,

[sk84]

For the same reason, we also have an internal vRA platform consisting of only one vRA appliance and 1 IaaS server. For our internal developer team.

In our case, the minimal setup is sufficient for this. But it is just not really a performance beast and I'm afraid if we had more than 5 developers making multiple changes a day, we'd have to switch to a distributed setup. :smileygrin:

[MRoushdy]

You will need to do this from scratch, as there;s no direct upgrade path from a minimal installation to a distributed installation.

[daphnissov · Accepted solution]

What I would probably advise in your case and what I've done for many customers is to create a distributed deployment but only single-legged using DNS records to simulate a LB. This entails 1 café appliance, 1 web, 1 manager, 1 DEM/Agent VM. You get more scale than minimal but are in a position to scale out and use real load balancers if you later choose. Granted, this is NOT a quick and simple process and you may want some PS help to do it, but it is technically possible to then convert this "HA-ready" deployment to a full on HA system.

[MRoushdy]

Yeah, that;s true, VMware recommends the same enterprise deployment with 1-node model. I've gone through many documentation,and still reading. The product is powerful and it does worth it.

[daphnissov]

Although I didn't call it out specifically, the implications of this design assume (and you had dang well better provide) proper and frequent VM-based backups of all components, not forgetting the IaaS SQL database. If using something like Veeam, for example, should one of the components fail and you have SureBackup configured, you can be up and running again in a matter of minutes. There's additional complexity there, but this is one way to mitigate such availability issues without going full-blown HA. As with all things YMMV and experiences differ...