Ok ok, I know. After all I said about using my Dell R410s, I went and bought new servers. They were cheap enough, and if you are in the market, I can recommend them. I bought 2 Dell R620s with 64gb of ram each, 2 Xeon E5-2670s, 2 PSUs, 4 SAS drives, iDRAC, and NICs all preinstalled, for $240 each. That is honestly a pretty good deal. The biggest reason I bought them was because of the price of RAM. I could not have realistically bought that much RAM to upgrade my existing machines, and as we talked about before, RAM will directly limit how many VMs you can run. Then there is the CPUs. Don’t get me wrong, the ones I have are fine for a few VMs. A bit slow, but otherwise OK. However, these older cpus also have some security vulnerabilities because they have not and likely will not receive any more security patches for bugs like Spectre and Meltdown. Making matters worse, these are the top of the line processors available for the R410, meaning there is no future upgrade path. Meanwhile, the E5-2670 processors are based on the, still old, SandyBridge architecture, but can be upgraded to a 2600v2 CPU based on the slightly newer IvyBridge architecture. It’s not a huge jump, but at least it’s there, and it is a pretty big jump from the R410s. I started looking at getting additional servers because I realized that I would eventually need to anyway in order to set up HA Proxmox. I wanted to buy 3 of these guys, but only 2 were available at the time, so I will add a 3rd server later. It may be something different, but who knows.

Another great advantage to upgrading to these newer servers is the additional upgrade to a newer iDRAC 7 Controller. This new controller has given me no issues whatsoever connecting to the virtual console compared to the iDRAC6. I am comfortably using it on my Linux workstation with no issues. And I am even using Java 1.8!

One big disadvantage to these servers however is the lack of ability to pass the hard drives through directly to the operating system. This is simply a limitation of the built in PERC H710 Raid controller. Ideally, you would use an HBA and let Proxmox handle the RAIDing of drives, which makes it far easier to transplant the drives to a new system or to replace the HBA if it dies. The only option with this controller is to create multiple RAID 0 arrays with 1 disk each, but that isn’t really a good idea either because if you try to move the drives to a new controller, they likely won’t work. Oh well… Unfortunate, but not a deal breaker. These are not storage servers, they are VM servers, and they will be backed up to external storage servers, so not the end of the world if something breaks. I have the option of using one RAID 0 array with 4 disks for maximum performance, which I am tempted to do, but I think I will set up a RAID 10 instead. I’ll get decent performance, but have some fault tolerance as well. Maybe I’ll switch this later once HA is set up, since in that case, everything will keep running even if the whole node dies.

I installed Proxmox onto this RAID 10 Array. Since I already created a cluster on my first server, all I had to do was join it with the second.