Dell R420

Hardware Enterprise Gear
21

There are lots of solutions that I would consider enterprise-grade, that are FOSS and very popular in the enterprise. Some of the most popular are TrueNAS Core, Proxmox, XCP-ng and “generic” KVM (oVirt or virt-man). For virtualization purposes, I only recommend the later 3. For just NAS and network file sharing, TrueNAS Core (unless you want to customize your own setup from the ground up). TrueNAS Core is great for what it does (being a NAS OS), especially if you are a home user.

Both TrueNAS Core and Proxmox have easy (too easy!) interfaces through which to configure ZFS pools

2020-11-20_16-11-1605883028
2020-11-20_16-11-16058830281381×516 39 KB

In the above, you would select disks. I didn’t do it through the GUI, because I had no idea if Proxmox would take the name assigned to the device (“sdX”), the name of the device (“ata-ST4000NM000A-…”), or the world wide name (“wwn-0x5000c500cd58…”). I preferred doing it through ZFS cli tools, because I wanted to make sure it uses the wwn and the pool stays the same even if I switch hardware. If I had any free disks, they would show up here.

The point is… both FreeNAS and Proxmox have easy GUI for management. But don’t rely on them. Use the CLI if you want to work in the domain in the future. For purely home use, they are good and extremely easy. I’m not sure about XCP-ng, but I don’t think XenOrchestra has GUI tools for ZFS (I could be wrong). But even if it did, I still recommend the CLI.

22

I was thinking about this, what if I made this server a dedicated TrueNAS server, and host all of my data (cloud, foundry, blog, media, etc.) on this device? I ask this because I’m having trouble finding a good nvme bootable solution within decent price tag. I’ve been kind of tempted to just buy two 250g usb sticks for the mirror array; leave pfsense on it’s current hardware; and build servers from there.

23

It will work, but 250GB is pointless imo. 16-32GB that’s max, just for boot os, that’s what I use sometimes.
But VM os images, that goes on SSD/HDD. There’s really good chance that 250GB stick will die while rebuilding and it will be slow af for VMs os. Also r/w operations will probably wear those sticks very fast.

So except os as bios extention theres no point really to going big.

1 Like
24

As misiektw mentioned, you only need 16-32gb for the boot device. I would argue, unless you want 99.98% uptime of your box, you don’t need to raid1 your boot/root. And even if you want 99.98% uptime, you still need to buy a UPS and a power generator, the later which doesn’t make sense for home users. I do however recommend an UPS for anything, be it desktop PCs or servers (they saved me lots of times and it’s how I made over 1 year uptime on my pfSense box and my ISP’s ONT and they have been especially useful when wfh, while my neighbors were in the black, I kept working like nothing happened).

I run my servers on 1 small ssd (between 120 and 250gb, with all of them being overprovisioned at least 30%, because I’m insane, I’m not even doing read-intensive workloads on them, let alone write-intensive) and backup the important files, usually /etc and /var (on Proxmox I only backup /etc/pve/qemu-server/ ). I saw people doing full system backups by copying everything under root, except dev, proc, sys, tmp, run, mnt, media (the last 2 usually being mounting points for external media anyway).

For OS that aren’t very customized by the user, like Proxmox and TrueNAS Core, I don’t think full system backups are necessary, because it’s easier and cheaper to just install fresh if your OS drive fails. But just for your ease of mind, do full system backups on your TrueNAS Core and only have it installed on 1 drive. Assuming you can’t find SSDs smaller than 64gb and provisioning them 7% (~58gb), and assuming you fill all that space up somehow, with 3 full backups retention policy, you sacrifice around 200gb from your 8tb zpool and you can also get a very cheap 500gb external hdd for cold storage backup, so your main OS data is safe. And that’s a worse-case scenario.

When giving those kind of data center at home advice, I always start laughing, because doing something like this in a production environment would be completely nuts, however, it just makes so much sense and saves a buck for home users, when availability is not a big issue.

1 Like
25

Yeah I agree pretty much with everything.

Just one more thing, if you have UPS, then redundant PSU is what I call nice bonus to have, because you don’t have shutdown even for battery replacement.

As for redundancy, all depends what your personal cash/time ratio is.
Because even with perfect backups it takes way more time to put it all back together on disk failure, than just replace drive.
That’s why even on my home rig I RAID1 almost everything. Only single drive instance is my WGC vm. Because I really don’t care if it bugs out.

(WGC - Windows Game Console ™ :wink:

1 Like
26

I’ll do you one better. I require an UPS. Yup. You can’t build your server without one. I won’t let you.

(All kidding aside, seriously make sure you have an UPS… otherwise you’ll hate yourself when something goes wrong.)

1 Like
27

Just last night around 3:20am I had a power outage, the UPS beeping woke me up, lasted for about 25min. Sure, being waken up in the middle of the night and hearing beeps for almost half an hour isn’t enjoyable, but I sure enjoyed having my computers protected against the nasty blackout. Reminder that I sleep near my LackRack (around 9 ft away). The loudest thing in it is my newest addition, a HP ProCurve 2910al 48 port switch (its fans are whiny, was thinking to change them with Noctuas). Once again, UPSes save electronics and uptime.

28

If you take the speaker from the ups and put a potentiometer on it, surely you can create a volume knob for your UPS since it’s right next to your bed!

Luckily for me there’s a floor of separation between me and the UPSes (… UPSs?.. whatever the plural is)

29

I could do that, but I want my warranty. :innocent: And I won’t be living in this studio apartment for long anymore, will move all my equipment to my mom’s apartment and I’ll move to New Hampshire.

30

Fair enough re: warranty.

I just can’t imagine sleeping with the threat of that annoying UPS beeping sound

1 Like
31

Yeah, UPS everything. A few crashes some years ago taught me that lesson. Total data corruption while I let something run, I don’t remember what it was now. But totally boofed whatever OS I was running. I think it was windows and woke to a BSOD after a power outage. Yeah, UPS everything.

1 Like
32

After a few installs on a thumb drive, different distros, etc. I just don’t trust it. I decided to go the m.2 route. I couldn’t find a free sata port on the mobo. But what I did find is a populated sata port for the cd drive, that I opted against when I bought the server. I decided to get a laptop cd sata converter and a 2.5 drive to dual m.2 sata. Two m.2s at 120GB. I think this will work out.

33

I have to disagree. UPSes are a common cause of power loss. In an area with generally reliable power, home users are likely to have better uptime without a UPS.

Practically all UPSes do naive switch-to-battery-and-pray self-tests periodically (Eaton units with ABM are the only exceptions I’ve found after many years of experience with many different brands, large and small). If the battery is in a bad enough state, this will drop the load. There are also plenty of situations where a sagging line voltage will cause the UPS to kick in and again drop the load if the battery is weak or the UPS is nearing overload, even while directly powered systems ride right through.

With dual-power supplies on two different UPSes, you’ve got much better odds of survival, but don’t underestimate the ability of UPS manufacturers to do something stupid, like APC’s perfectly synchronized self-tests EXACTLY two-weeks-to-the-second after utility power last came on. The sudden drop and surge in electrical demand can trigger other UPSes to switch off line power briefly, and even trip breakers in extreme cases.

At a minimum, if you’ve got a single UPS powering anything important, replace the battery after no more than 3 years while it’s still working well, BEFORE you see any issues. And use only reliable name brand batteries.

34

Then don’t buy an el-cheapo UPS?

35

Well, that’s one mystery explained for me by accident, that I don’t need to google after :slight_smile:

Now I know how to schedule test on my home APC, so it doesn’t start beeping at some stupid hour when I’m sleeping, like in the middle of the day :wink:

1 Like
36

APC is overpriced sh*t if you ask me. I had at my last workplace around 50 APC desktop units, 50% loaded at most, almost 20 of them failed between 3 weeks and 9 months after warranty ended. Eaton is a nice brand, if the 9130 was quieter, I would have kept it home for my LackRack Lab, but it felt like a vacuum cleaner staying near it (I powered it off and brought it back the next day, because I couldn’t sleep because of it).

I got generic UPSes, one protecting my DVR, router and RPi Wireguard VPN to it at a remote location, another one at home powering my pfSense box (AsRock J3455M), ISP’s ONT, my Zyxel router in AP mode, a Pi 4 8gb (main desktop) and another J3455M mini-server (used to be my old desktop). Just bought another one of these to protect my used server soon-to-be NAS and VM Host and another PC. I had so many brownouts lately, I can’t believe it.

I had an old APC BackUPS 300, lasted me 20 years (only ~10 years of usage). I had a newer model (Back-UPS CS 500) that lasted 5 years of usage, with battery changed, before it blew in my face (and it had a terrible coil whine). I don’t trust newer APC UPSes. Generic ones, if they aren’t complete sh*t inside, should be just fine. Whenever I had a brownout, the UPS would kick-in and supply my computers with enough juice for a few seconds until the power got stable again. Additionally, I mentioned above the blackout at 3 am, where this generic puppy saved my computers from imminent poweroff.

As for UPSes triggering a brownout, if your other UPSes don’t switch to battery fast enough, then you got a bad brand. And if your UPS is tripping the breaker, you are sucking too much juice from the wall. My old APC BackUPS 300 kept my pfSense box, ONT and Zyxel router for 1 year up (around 400-something days of uptime BSD Challenge ). Nowadays this APC (and also the one that blew in my face) wouldn’t switch to battery fast enough (and I replaced their batteries) and would restart my systems, so I removed them.

tl;dr: you can buy cheap unmanaged UPSes, don’t get the cheapest of the bunch, get more power than you actually need, so you don’t overload them, don’t keep all your devices on just 1 UPS and one power line, split the load and always change batteries 2 or 3 years apart. You wouldn’t keep using your oil in your car, so why would you keep using an old battery in your UPS?

37

aka don’t buy trip-lite. They’re absolute trash. I have around 25 of them at work I’m trying to phase out right now. They cause more harm than benefit. Garbage.

38

There’s no way that this is accurate. MOST individuals experience at least one 30 second power outage event every 3 to 4 months. I’ve got 213 days uptime on an unraid server (with 180 or so days prior to that - only needed to shutdown to physically move its location). I’ve definitely experienced at least 4x 30 second power outages during those times, all easily mitigated with an UPS with no issue. In all the places I’ve lived (all reasonably populous non-rural environments), I’ve never experienced a situation where the power just NEVER goes out.

You know what is a more common source of power loss? Power loss. 100% guaranteed system down when your power goes out.

This is like saying cars are unreliable if you don’t put gas in them. This isn’t a reasonable criticism of UPS, but rather a criticism of unreasonable behavior. Change the batteries every 2.5 years on a schedule (or sooner if your power is more prone to frequent drops). And obviously set-up your UPS to power-down your system if you’re experiencing a power event that exceeds 50% capacity so you never drain the battery down to 0%.

I promise that you won’t get 2.5 year uptime WITHOUT an UPS in any house in America.

3 Likes
39

If you’ve got a source for that, I’d love to see it.

Cars have accurate gas gauges. It’s not possible to have an accurate battery gauge. It’s a bad analogy, too. More like you get suddenly ejected from a working car if you haven’t cleaned the seat in a while…

Quite a few cities go that long between power events. Of course some areas are unlucky and bring the average down.

Here’s the list: https://www.eia.gov/electricity/data/eia861/

40

I was referring to quite a few UPSes across a data center, though once or twice I’ve seen it in private homes with several large pieces of equipment on UPSes blowing the main 100A breaker. Sudden synchronized drops then spikes in electrical demand are no good.