whipdancer

Unless your UPS has a feature to automatically turn on when power is restored - you don't.

Unless I can find factual evidence of it negatively impacting my use, I don't worry about it.

Sent you a PM.

I assumed you were referring to the physical slot (you only have a 4x physical slot). Assuming that is the case, then an adapter will work. If you have the physical slot size, but it run at 1x (or something below optimal) it will also work. Both cases, the result is that you lose bandwidth. Here's an example - http://www.cryptocables.com/wp-content/uploads/2013/04/1x.jpg (just something I found, not saying buy this one or from this vendor).

I've never seen a PSU with 20 sata power connectors, either. At my worst, I was running 24 drives using a number of power splitters (1-n power splitters/adapter cables) on an 800w PSU. I just tried to split them between the 2 rails the sata connections were on. I've since gone for higher density and am actively reducing drive count.

I've only ever dealt with 8x and 16x HBAs. You can use an adapter to fit an 8x (or 16x) in a 4x slot. You lose bandwidth but it will still work.

Are you talking about power splitters? Regardless, you will need enough power to deal with the spin-up power requirements. A relatively simple calculation to estimate. One reason you may want a more expensive PSU is to have separate rails. When powering that many drives, you want to make sure you don't starve something else because you have momentarily heavy demand.

You can start here.

This card runs hot. I was warned about it when I bought it. I run dedicated fans for it. When I asked about monitoring it, I was told just as @JorgeB mentioned - they weren't aware of built-in way to monitor it.

After I preclear it, assuming it passed, yes.

I have carpet downstairs, so I ran Cat6 from the router that is physically next to the modem tucked under the carpet, along the baseboards from the back of the house to the fron t of the house. Previously (different rental), I can all my Cat5e thru the attic and did my own drops in the walls I was interested in. Then all I had to do was add a new "outlet" to the wall for the ethernet cable. Why not have them run it through the attic?

Can you run wiring yourself (I do that even when I'm renting)?

I estimate ~9w per mechanical drive, with room for about 10 drives per plug on my PSU. I use a relatively high quality 700w power supply to power 18 drives currently, including 15 mechanical, 2 SSD and 1 optical.

You can find extra long power and sata cables on Amazon. You can also find sata and/or power extenders on Amazon. You can also find 6 pin cables for your PSU that end in 4 pin molex connectors (old school power supply plugs) and you can then use long 4 pin to sata power cables. The schematics are not that difficult so you could also roll your own. You can see the case I use (for now) in my sig.

I understand that and have been in that situation before. I once went over 2 weeks with my missing drive in emulation mode. Never again. I always buy a spare. In fact, when I buy a drive, I'm replacing my spare. The drive that was my spare will be going into my server. I also put every drive I buy through a full stress test. I use the preclear plugin (3 full passes) if I can connect it to my server or one of the other drive burn-in scripts I've found if I have to do it some other way.

Run an extended smart report. That's a common practice, so I'm not surprised. Since you're running parity, you can remove the offending drive, install a replacement and allow it to rebuild parity. You'll be fine as long as you don't lose another drive.

No. I think it’s likely you got 2 bad drives. I’d look at the date code info on all 4 to see if they are all from a single batch/lot.

I’ve got 2 of them, and 2 x300. No issues. All have over 25k hours now. It’s a crapshoot - doesn’t matter the brand.

That expansion unit retail is over $2k. I would rather pay to have this case printed (about $800) - or see if I can print it myself. One or the other will eventually happen as I'm actively downsizing the physical dimensions of my server - currently the gigantic full tower Antec Nineteen Hundred case.

At $100 a drive... to me... never. But I'm not searching for that last, extra performance bump either.

What's the cost difference?

Since the wording might be confusing, I'll say it a different way. Parity drive(s) must be equal to or larger than any other single drive in the array. If your parity drives are 22gb and 20gb, then the largest any single drive in the array can be is 20gb.

No VMs. I ask my server to do enough to keep it busy already (20-ish docker containers) and it's just running a pretty ordinary desktop cpu. If you want to run a vm, you would probably want to run it on an SSD. I personally wouldn't spend the money on a high-end drive for a VM because I don't expect to see a difference between a Gen4 x8 and a Gen3 x8 NVME running through a hypervisor. I've never measured so, I could certainly be wrong. It's not something I would even worry about - I'm all about the value curve, not the performance curve.

I always keep a spare the same size as my largest drive. If I upgrade my largest size, the old spare goes into my server (and I make sure I order 2x of the new larger size or it doesn't go into my server until I have a spare).

Why buy something like that when there are so many other options? I can't speak to compatibility of the card you linked to nor would I want to consider the implications of running 16 drives over that connection. Simply put, there are any number of HBA's that absolutely will work and cost less.