NEWBIE - First build mITX, ITX Home Server

[mijki]

I don't want to be cheap about anything, I want a reliable system. On the other hand, I'm in Central Europe, so I do have not too much money (1200 - 1250 USD with drives) and the market is a bit different than in the US for example.
For my, AMD builds I got two suggestions, I should get a dGPU for Plex/Jellyfish and an HBA card for the ESXi / Proxmox. And instead of 4 * 2 TB disk, I should use 2 * 18 TB Exos HDDs.
Do you think these are necessary for the build or not?

@WI_Hedgehog are you talking about rack servers right?

 

[WI_Hedgehog]

I completely appreciate living in Budapest stands in stark contrast to Wisconsin, because I have lived in one of your neighboring countries and realize the difference in available resources. Such as history and architecture, you have an abundance whereas ours is undervalued and left to wither.

In my estimation you are looking at three discrete, inexpensive yet completely reliable systems that require very little maintenance while providing rock-solid services:

1. File storage. Fast, simple, highly reliable. Like a hammer, it does the job it is intended to do. The main purpose is storing data from your research group, but on the side it stores private data and videos. This system doesn't need much CPU power, or excessive amounts of RAM, but security does need to be maintained with simple, straight-forward security patches applied over time. TrueNAS is excellent for this.
2. A Virtual Machine server with RAM, CPU power, local SSD storage for the virtual machines (not the data, the VM), and a fast connection to the NAS. Spool up a VM for your backups so if, somehow, your NAS is hit by ransomware (or other unexpected catastrophic event) you still have clean backups. Server-grade hardware is again the safe bet so things run consistently smoothly, especially if running a web server. This is quite possibly where the Internet Portal/Portals are hosted, in secure VMs. Proxmox is a sound choice. Running Windows 11 simulators will take some computing power and RAM, this machine should have it.
3. This one is optional, but a fun/experimental server isn't a bad idea. Run Proxmox on it and a VM for Jellyfin (Plex is dying). If you need to make changes to your main Proxmox server, test them out here first so if you seriously break something the worst that happens is your co-worker's stream of Yes, Minister is dead for a day or two and they become far more productive. This seriously affects VMs when security updates roll out, because breaking functionality here on the Test Server doesn't affect your Production Server. Plus, if you need to upgrade your Production Server (or someone breaks it) there's not a lot of stress on you because the Test Server can run the VMs (not as fast, but the staff is still productive). And, if a gaping security hole is found in Jellyfin or whatever other non-essential software you have running it doesn't affect your Production Server.

I say this because adding complexity to a system exponentially increases the difficulty in keeping it running, and although the up-front cost of three systems is slightly more the effort and mid-term costs are far lower. Plus, you are more valuable because everything is highly reliable.

BUT, this advice is not for everyone. If you love drama, please jam it all on one server, and add in a Virtual Minecraft Server so you and your co-workers have some LAN Gaming time on Fridays. DO NOT use ECC memory, but DO use RAID-0 striped arrays for blistering HDD speed. There's no need to verify your backup strategy is working properly, well, don't bother with backups and spend your time in Minecraft instead, because when this starts to fall apart it's going to be in bigger and bigger chunks, and your drama will unfold nicely.

Personally, I like to spend my days in my office, passed out in my very comfortable chair, basking in the sunshine streaming through the windows. They don't pay me any less when things run flawlessly, in fact during reviews they generally opt to pay me more. But that's not the job for everyone, I just happen to like it.

 

[mijki]

I completely appreciate living in Budapest stands in stark contrast to Wisconsin, because I have lived in one of your neighboring countries and realize the difference in available resources. Such as history and architecture, you have an abundance whereas ours is undervalued and left to wither.

In my estimation you are looking at three discrete, inexpensive yet completely reliable systems that require very little maintenance while providing rock-solid services:

1. File storage. Fast, simple, highly reliable. Like a hammer, it does the job it is intended to do. The main purpose is storing data from your research group, but on the side it stores private data and videos. This system doesn't need much CPU power, or excessive amounts of RAM, but security does need to be maintained with simple, straight-forward security patches applied over time. TrueNAS is excellent for this.
2. A Virtual Machine server with RAM, CPU power, local SSD storage for the virtual machines (not the data, the VM), and a fast connection to the NAS. Spool up a VM for your backups so if, somehow, your NAS is hit by ransomware (or other unexpected catastrophic event) you still have clean backups. Server-grade hardware is again the safe bet so things run consistently smoothly, especially if running a web server. This is quite possibly where the Internet Portal/Portals are hosted, in secure VMs. Proxmox is a sound choice. Running Windows 11 simulators will take some computing power and RAM, this machine should have it.
3. This one is optional, but a fun/experimental server isn't a bad idea. Run Proxmox on it and a VM for Jellyfin (Plex is dying). If you need to make changes to your main Proxmox server, test them out here first so if you seriously break something the worst that happens is your co-worker's stream of Yes, Minister is dead for a day or two and they become far more productive. This seriously affects VMs when security updates roll out, because breaking functionality here on the Test Server doesn't affect your Production Server. Plus, if you need to upgrade your Production Server (or someone breaks it) there's not a lot of stress on you because the Test Server can run the VMs (not as fast, but the staff is still productive). And, if a gaping security hole is found in Jellyfin or whatever other non-essential software you have running it doesn't affect your Production Server.

I say this because adding complexity to a system exponentially increases the difficulty in keeping it running, and although the up-front cost of three systems is slightly more the effort and mid-term costs are far lower. Plus, you are more valuable because everything is highly reliable.

BUT, this advice is not for everyone. If you love drama, please jam it all on one server, and add in a Virtual Minecraft Server so you and your co-workers have some LAN Gaming time on Fridays. DO NOT use ECC memory, but DO use RAID-0 striped arrays for blistering HDD speed. There's no need to verify your backup strategy is working properly, well, don't bother with backups and spend your time in Minecraft instead, because when this starts to fall apart it's going to be in bigger and bigger chunks, and your drama will unfold nicely.

Personally, I like to spend my days in my office, passed out in my very comfortable chair, basking in the sunshine streaming through the windows. They don't pay me any less when things run flawlessly, in fact during reviews they generally opt to pay me more. But that's not the job for everyone, I just happen to like it.

Hi,

Firstly I'm sorry if my questions bother you.
So as you suggest, I should separate all these systems - at least the VM and the NAS - on completely different hardware?
Or is it doable if there is one hardware with a supervisor (ESXi, Proxmox) and they are running in containers separately from each other moreover they can use separate HDDs/SSDs also? So if the VM brokes, for example, it wouldn't affect the NAS or would it?

 

[ChrisRJ]

So as you suggest, I should separate all these systems - at least the VM and the NAS - on completely different hardware?
Or is it doable if there is one hardware with a supervisor (ESXi, Proxmox) and they are running in containers separately from each other moreover they can use separate HDDs/SSDs also? So if the VM brokes, for example, it wouldn't affect the NAS or would it?

My view is the following (it may sound condescending, but is not meant this way and comes from having seen too many people shoot themselves in the foot with similar endeavors):

If we are talking about a scenario that is not purely a playground system, the mere fact that you are asking those questions now, means that you should keep things separate. Think about a situation where something has gone wrong. Unless you know and fully understand your system and how its different layers interact, you cannot quickly resolve the situation, because you depend on others. Those, in turn, can only guess what is going on, because there are tons of variables they don't know. And without sitting in front of the system, there is a high risk of someone giving wrong advice, simply because they don't know a tiny detail that changes everything.

It is a bit like with a car failing. The "problem statement" that a normal drivers can give to the repair shop, will almost certainly not be helpful beyond describing roughly where the "anomaly" occurred. But if the driver is a trained mechanic, they have observed things like "in the past the engine start took 2 seconds longer, if the temperature was below 5 C and a lot of humidity was in the air".

For me it all comes down to complexity and how it affects operation in case of a failure. I am certainly biased, given my enterprise software background. If something goes wrong in that context, no lives are at stake, but loads of money. Therefore avoiding any complexity that can be avoided, is one of the biggest levers to make things safer. Lastly, think about what would happen if you were on vacation and someone else would need to fix things.

Hope that helps.

 

[mijki]

I trust you guys, I have never worked with home servers before. So then I will slowly build up my systems, step by step.
Then I will need some time to rethink what I should get if I only want to run Proxmox VM or bare metal TrueNAS Scale.

Then the webserver could run on my Raspberry Pi3 for example.
And I will found out something for the VM, I think Oracle has some FreeTier solution for that.

 

[ChrisRJ]

I trust you guys, I have never worked with home servers before.

Often there is not simply right and wrong. Instead, a lot of nuances and various levels of grey exist. They depend on the exact details of the use-case, the previous experience of people, and also how good or bad a day they have. Whenever you get contradicting advice, it is worth getting to understand how this can be.

So then I will slowly build up my systems, step by step.

Assuming you have the time, that is always good. Plus you will learn a lot, simply by following the forum and reading.

Good luck and enjoy!

 

[Etorix]

TrueNAS baremetal is safe for your data. That's a good start. From that, you may experiment with VMs and/or containers and eventually migrate the web server from its initial Pi to the NAS.

 

[Whattteva]

I trust you guys, I have never worked with home servers before. So then I will slowly build up my systems, step by step.
Then I will need some time to rethink what I should get if I only want to run Proxmox VM or bare metal TrueNAS Scale.

Learning is a journey. If you look at the systems in my signature, you can see how the systems evolve with my understanding and knowledge level basically from gamer gear to baremetal rack server to a hypervisor. That entire evolution didn't happen in a day and happened over the last 10 years. Take your time and enjoy the journey.

 

[WI_Hedgehog]

Hi,

Firstly I'm sorry if my questions bother you.
So as you suggest, I should separate all these systems - at least the VM and the NAS - on completely different hardware?
Or is it doable if there is one hardware with a supervisor (ESXi, Proxmox) and they are running in containers separately from each other moreover they can use separate HDDs/SSDs also? So if the VM brokes, for example, it wouldn't affect the NAS or would it?

Your questions are welcome, we are here to help you be successful with as little stress, cost, time, and frustration as practical. We want you to succeed and do well.

TrueNAS can run well under Proxmox, and there are good reasons for doing so, such as your Virtual Machine Manager (Proxmox) has incredibly fast access to your Network Attached Storage (TrueNAS), and your hardware resources are more efficiently allocated (CPU, memory, storage).

HOWEVER, getting this to work well on one system is not easy. When any software breaks, fixing it will be difficult and sometimes seem impossible, though it will be possible given enough time, experience, and knowledge.

Think of it like this: Every relationship affects every other relationship. A group of 2 people each have one relationship, or two relationships total, and they either work together or they don't. A group of 3 people each have one relationship with 2 other people, or six relationships total, and one or more people may have inter-personal conflicts that affect the third person. 4 people have 3 relationships each = 6 total. Here's a table:

Number of People	Number of Relationships (n-1)*n
1	0
2	2
3	6
4	12
5	20
10	90
20	380
50	2450
100	9900

Relating that to your system, how many processes are active at any one time? Not all of them talk, but if you need them to and for some reason they're not talking (or are but not understanding each other)....finding the source of one problem is difficult, finding the sources of 3 concurrent options is very difficult. Imagine a simple data corruption issue: is it a drive, cable, controller, virtual controller, memory, shared memory issue, lack of memory issue, access conflict, pass-through conflict, software bug introduced in an update--and which piece of software...

You want to acquire the skills in each package (Proxmox, TrueNAS, your virtualized applications) separately, so you can more easily isolate the source of issues (including security holes). Once you're a relative "expert" in each area, then adding the complexity of everything on the same system with virtualized hardware, pass-through hardware, resource sharing, etc. will be a new challenge, but manageable.

 

[mijki]

Thank you guys for everything, this forum is one of the most helpful where I ever was.

I found a used config on eBay for 600 pounds, and I want to ask if this will be good to be a starter config?
- Supermicro X10SRL-F LGA 2011 motherboard, with original packaging and all accessories.
- Intel Xeon E5-2680 V4, 2.4GHz 14 cores CPU.
- 128GB DDR4 ECC RAM, (8x16GB).
- Noctua NH-D9DX i4 3U CPU cooler, with original packaging and all accessories.
eBay link

 

[mijki]

Or I found this, this would be better for a 1st Home Server doesn't it? It is an mATX board, so it can fit in the Node 804 and the mobo has 10G RJ45.

• Supermicro X10SRM-TF motherboard with IO plate
• Intel Xeon E5-2680v4 CPU 14 core 28 thread at a base 2.4GHz, tons of compute! (2690v4, 2680v4 and others on hand for upgrades)
• 128G RAM 4 x 32GB Sk Hynix DDR4-2133 ECC REG RAM
• Samsung 128G NVMe (perfect for a boot drive)
• Large heatpipe cooler

eBay link

 

[Etorix]

For a converged NAS with lots of VMs/apps, either is fine. For a simple NAS, plainly serving files over SMB a lower CPU with less cores at higher clocks would do better, but that's easy to arrange.
Our network experts advise against 10GBase-T "if you can". I guess, though, that the X10SRM-TF is a case of "I got copper 10G on-board, so I will use it anyway".