Homelab downgrade, sanity check needed!

[DRAGONKZ]

Hi Guys,

I’m weighing up the idea of downgrading my homelab and wanted to get some opinions before doing so.

I’ve currently got a 42RU Dell rack, 17FP Dell kmm, 2 x Dell 8024F switches in a stack, 2 x Liebert 1kva gxt4 UPS, Synology RS815+ with 4 x 3TB WD Red drives and a 4 x HP DL20 Gen 9 servers in a VMWare vSAN cluster.

The HPs are almost fully optioned, with the main bits being each has 64GB RAM, a HP H240 HBA, 1 x 400GB HP branded HGST HUSMM80 SAS SSD, 1 x 600GB HP branded Intel S500 SATA SSD, and a Dual port 10Gb HP 546SFP+ NIC.

I built my lab to help with study for VMWare VCP and VCAP certs, and to also build a mini enterprise environment to mirror my work one, to allow me to upskill and test.

My lab has helped me achieve the above, but I’m at the stage now where my usage of it isn’t reflecting the investment in it, and the sheer bulk of it also makes it hard to house.

I never planned on having this exact set up, it’s just organically grown in to this for better or for worse.

I’m also at the stage where if I want to continue to grow it I start hitting limits with licensing and have to look at spending thousands just to get more RAM!

I still want a lab, I’m just at a point now where I’m thinking I may be better off downsizing to a single beefy server and lose some of the redundancy and complexity in the process.

I’m currently considering a single Dell R720 server, with 2 x e5-2690 v2 or e5-2697 v2 CPUs, 384GB RAM, and 1 or 2 Intel P3600/P3700 PCI-E SSDs.

I’d also look to get some 4-6TB NL SAS drives and move my backups from the synology to these.

I’d essentially go from a 1/4 rack of gear to a single 2RU server...

Has anyone gone through a similar process, and if so, were there any main negatives you experienced?

Any comments/thoughts greatly appreciated

 

[MiniKnight]

Was V2 where nested virtualization got good? I thought it was V3?

If you want to go crazy, you can get a 4-socket cheap these days. Nested virtualize an instance onto each CPU and you've got a 4 node cluster in a single machine.

 

[whitey]

If you need to downsize and do not intend to run vSAN you can certainly come down to a single node or two node (think vSAN supports this) setup. I recently went from a 3-node setup to a 2-node but mainly for an architecture update and to save a lil bit o' power. With a beefy host w/ 192-384 GB of memory and a decent core count you can still do a TON w/ that. Spin up nested hypervisor hosts and still reap the benefits of a multi-node cluster w/ most all the advanced features still albeit a lil' less redundancy/resiliency. Just make sure you get recent processors that support vt-X/vt-D but pretty much anything w/in the last 5-7 years will inherently have those feature sets.

What do you have to lose, you can always scale back up, I'd start by getting rid of that monstrosity of a rack that is WAY more than you actually need, no need for a physical KVM as well, IPMI virtual console and media are insanely useful. Do you REALLY need that dual switch stack as well, if you like those 8024F's I vote keep one up, tell me this, when was the last time you actually experienced a switch failure that warranted dual switch setups in a home lab...I'd venture to guess next to never. Keep one spare or sell it off to recoup costs on a new beefcake host, could also go Xeon-D into 1U chassis/cases and REALLY slim down pwr, environmentals, and space. :-D

GL buddy, whitey

 

[Evan]

vSAN works in 2 modes + witness , witness can run on say an intel NUC drawing less than 10 watts.

For always on I keep thinking I could live with 1 node (spare not always on hardware even, maybe auto start weekly for backup images) but then I keep coming back to the 2 nodes for redundancy.
vSAN maybe, Hyper-V with storage replica maybe, or S2D. A proxmox or KVM solution would work.
StarWind vSAN would work I guess.

Or maybe just 2 nodes where you have application redundancy, eg failover with pfsense rather than single VM’s restarting on other node during failure.

 

[Aestr]

As mentioned above this is very possible from a technical perspective, using many different means. I think you need to sit down and figure out what your requirements are before you start buying and selling. For example if redundancy is something you need at a host level (rather than just good backups) build that in, and if it's something you just want to play around with you can nest hypervisors as mentioned above.

I'm also not sure about what your goals are here and I think that will help you choose a path. Are you looking to reduce the physical footprint, save on power consumption, recoup some of your labs costs or shift to more recent gear? Many of these can coexist but you'll find you have to strike a balance. Newer gear, space saving and power savings all work well together, but in many cases you'll end up spending more to get there.

Make sure you meet your absolute needs as well as the nice to haves that are going to bother you on a regular basis. I've done this seesaw several times now and the only time I've really regretted the choices was when I cut deeper into the nice to haves than I should have. I quickly ended up going back to a level had the right mix of function and fun.