Nothing beats a good use case. Like Microsoft, we “eat our own dog food”. For us, that means the servers running this forum, our website, our customer portal, etc. are located in the same racks as customer servers. Our Proxmox cluster is the same kind that we sell. It’s a custom cluster, but it’s still the same that we’d sell you.
Our cluster is a simple 3-node cluster of identical servers. We don’t actually use shared storage - it’s all local. But we utilize ZFS replication and replicate our critical servers between all three nodes at least every 5 minutes. Our nodes are interconnected in a 10 gigabit fiber network as illustrated in the following diagram:
So why don’t we use shared storage or even Ceph? Well, we’re budget conscious, but the main reason is we architect our infrastructure with a simple philosophy: Keep It Simple, Stupid! KISS is what we live and work by at NodeSpace. We’ve used Ceph before, but we’ve seen it implode really fast. We’ve used shared storage before - and likewise, we’ve seen it implode. The more complexity we can remove out of a situation, the better. Also, how can we prove to potential clients that we can build something within their budgets if we use top-of-the-line hardware and setups? Besides keeping it simple for our own team (which means if there is downtime, it’s faster for us to get resolved), it makes it easy for outsiders to come in and help us fix issues because they understand our infrastructure at first glance.
Besides our cluster, we keep a single Proxmox VE node available that hosts the non-essentials. Basically, if it’s a Tier 2 application or lower, it goes on our standalone-node where HA and redundancy are not important. All of our Tier 0 and Tier 1 applications and websites reside on this cluster.
To manage our cluster and individual node, we utilize Proxmox Datacenter Manager. While PDM is in alpha stage, we find it extremely beneficial. If we need to move a VM from the standalone node to the cluster, we move it with a single click.
We also utilize Proxmox Mail Gateway as our antispam filter and mail gateway. A lot of our secondary domains utilize PMG as their MX. For nodespace.com, all mail comes into Office 365 (for now) and the majority of our mailboxes don’t reside there - like our support mailbox. From Office 365, the emails are sent to our PMG server and from there, are passed onto our other mail server.
Finally, we back everything up to a Proxmox Backup Server with over 20 TB of available storage. We can backup our entire environment there, and we do. Prod and dev all get backed up daily.
In case it’s not evident, we 
Proxmox!