v3 Build Peer Review

[SaltyCoffee]

This community has been such an amazing resource on my TrueNAS journey. It would be malpractice not to run this build by the collective brains here. I'll try to be as complete as possible to minimize the need for questions (but all are welcome). I'll only request to avoid the rabbit hole that is the subject of ITX builds (which this is).

The build:

• Mainboard (Intel NIC, 4 SATA onboard, 3 M.2, 64 GB max RAM): ASRock B550 Phantom Gaming-ITX/ax
• CPU: AMD Ryzen 5 PRO 2400G
• RAM: 2 x Kingston Server Premier DIMM 32GB, DDR4-3200, CL22-22-22, ECC
• HBA: LSI 9211-8i
• SSD (cache): 2 x Patriot P300 SSD 128GB M.2 NVMe PCI Express 3.0
• SSD (os): SATA coming over from v2 build
• HDD: 5 x Seagate IronWolf Pro NAS HDD +Rescue 22TB
• PSU: EVGA SuperNOVA GA 550 550W ATX
• Case: Jonsbo N1
• Plus Noctua fan and cooling block, and DeLock cables.

Now, just to get out in front of what I know is coming regarding airflow in my preferred case, I am going to monitor internal temperatures and see how it goes. If I find the Noctuas aren't getting the job done, I'll just get the FD Node case and be done with it. Also, I don't know if 256 GB SSD cache is overkill or not, but for 20 eur, you're free to convince me it's a bad idea (really, I'd like to know if it is).

I am also considering the ASRock Rack X570D4I-2T, but the only real advantage I can see with this board, apart from freeing up the PCIe slot (which would be used to regain the lost M.2 slots) w 8 SATA onboard, is the 100BaseTX NIC, versus the 2.5G on the Phantom. I'll only be PLEX streaming to max 3 or 4 people at any given time (and rarely, mostly just myself at home). This machine won't be doing much more than that. Not sure the 500 eur premium is worth a little wider network throughput that, if my assumption is correct, I won't be using really anyway.

But, that's the program. Would really love to hear why it won't work lol (letting me know about good choices is welcome too where appropriate, but not expected)

TIA

EDIT: Updated links to english

 

[artlessknave]

gaming motherboard is a poor choice for TrueNAS. not recommended.
asrock rack is a server motherboard. while less popular here in the forums, it's a far better choice. the price, however, does suck. *seriously* consider looking for a better option. Intel is still king in the server space.
I dont know your region, but there should be better prices available, particularly in used server hardware, which lasts basically forever and is generally fully supported. for example, the x9scm/e3-1230v2/32GB ram ebay combo is VERY difficult to beat, and will serve you just as well, for probably about the price of that gaming mobo. (your chosen CPU is only 2000 passmarks stronger than an e3-1230v2. for a server, newer is often not better.

SSD (cache): 2 x Patriot P300 SSD 128GB M.2 NVMe PCI Express 3.0

64GB RAM MIN before even considering l2arc. for most people there is no point, all it's likely to do is use up the writes on your drives and either do nothing or slow you down.

ASRock Rack X570D4I-2T, but the only real advantage I can see with this board

IPMI is a massive advantage.

versus the 2.5G on the Phantom

my understanding is that 2.5g is poorly supported, particularly on core. generally, either 1gbe is good enough or get 10gbe.

 

[SaltyCoffee]

Hey thanks for the input! Comments below.

gaming motherboard is a poor choice for TrueNAS. not recommended.
asrock rack is a server motherboard. while less popular here in the forums, it's a far better choice.

It makes sense. I'm curious to learn the why? I wasn't looking for a gaming main but that's what is out there in the ITX FF that isn't at an outrageous price. Not to say I'd be opposed to shelling out the cash if there was a tangible advantage to my use case, hence my desire to learn more.

x9scm/e3-1230v2/32GB ram ebay combo is VERY difficult to beat

No argument there. Sadly it won't fit in my preferred case.

64GB RAM MIN before even considering l2arc. for most people there is no point, all it's likely to do is use up the writes on your drives and either do nothing or slow you down.

Fortunately I've provisioned 64GB RAM, so it sounds like I'm in the game! But if my use case won't benefit, then I guess I could pass on cache drives. I need to learn more about it I suppose. I just assumed cache = performance.

IPMI is a massive advantage.

Certainly, altho I'll be 10 feet from this thing most times, so I have no need.

my understanding is that 2.5g is poorly supported, particularly on core. generally, either 1gbe is good enough or get 10gbe.

I did come across something like that, but it seemed to me to be a driver issue w Realtek. Haven't read any problems w Intel 2.5g NICS. I'd be interested in any info you might have about it.

Thanks again for the insight.

 

[SaltyCoffee]

A quick look at viable alternative mainboard/cpu combos, taking your advice into consideration, turned up these as the most cost efficient:

• Supermicro X12STL-IF
• Intel Xeon W-1250, 6C/12T, 3.30-4.70GHz

This option tacks on about $450 to my original total while maintaining compatibility with the rest of the build.

EDIT: Seems involving eBay can reduce the differential to around $200. That is in the ballpark of reasonable.

 

[artlessknave]

It makes sense. I'm curious to learn the why? I wasn't looking for a gaming main but that's what is out there in the ITX FF that isn't at an outrageous price. Not to say I'd be opposed to shelling out the cash if there was a tangible advantage to my use case, hence my desire to learn more.

IX builds truenas as a storage appliance built for enterprises on enterprise hardware for enterprise uses tested mainly on enterprise hardware to ensure their enterprise can be supported reliably, but releases the image to the public.
as such, it works best and most reliably on enterprise grade gear, with workstation gear being, while not exactly "supported", usually decently reliably. this tends to mean that "gaming" boards have issues, like using realtek NICs, having audio chips that could cause conflicts while being *utterly useless*, etc.

But if my use case won't benefit, then I guess I could pass on cache drives. I need to learn more about it I suppose. I just assumed cache = performance.

zfs cache is in RAM. both write and read cache are in ARC. that's why it's called l2arc. there is no write cache on disk. this confuses many people who are used to something like unRAID, which is constructed differently. the best time to add l2arc is after your setup is running and you have stats that show you would benefit from it.

No argument there. Sadly it won't fit in my preferred case.

that was an example. that board is old in the tooth but rock solid (beyond the java console being a pain)

No argument there. Sadly it won't fit in my preferred case.

your preferred case is not really a good choice for TrueNAS, due to the extremely limited selection of TrueNAS compatible boards that will fit.
that said, the TreuNAS minis use mITX boards. they have alternated between Asrock and supermicro boards for them.

also, have you checked out actual TrueNAS minis? if they are available they could be an option worth looking at.

 

[SaltyCoffee]

IX builds truenas as a storage appliance built for enterprises on enterprise hardware for enterprise uses tested mainly on enterprise hardware to ensure their enterprise can be supported reliably, but releases the image to the public.
as such, it works best and most reliably on enterprise grade gear, with workstation gear being, while not exactly "supported", usually decently reliably. this tends to mean that "gaming" boards have issues, like using realtek NICs, having audio chips that could cause conflicts while being *utterly useless*, etc.

Definitely intuitable, that. Although you provide some satisfying possibilities. I'm curious if anyone's done a benchmark test to show the difference on the ground. If such a thing is even possible. I'd be very interested to see.

zfs cache is in RAM. both write and read cache are in ARC. that's why it's called l2arc. there is no write cache on disk. this confuses many people who are used to something like unRAID, which is constructed differently. the best time to add l2arc is after your setup is running and you have stats that show you would benefit from it.

I was reading up on this, and apparently it has to do with how ZFS handles arc evictions to L2. Pro tip waiting for performance data, thanks!

your preferred case is not really a good choice for TrueNAS, due to the extremely limited selection of TrueNAS compatible boards that will fit.
that said, the TreuNAS minis use mITX boards. they have alternated between Asrock and supermicro boards for them.

also, have you checked out actual TrueNAS minis? if they are available they could be an option worth looking at.

There certainly is not. A dearth, to put it mildly. I suppose the bright side is that it cuts down on research time. But that's my quirk here, I want it to be as beautiful as it is functional, so long as I can maintain functional above board.

Regarding iX hardware, I have, in fact. Their pricing leaves much to be desired, if I'm perfectly frank. I suppose their premium is aimed at folks who can't be bothered to put in the sweat equity. But I actually find it fun, so for me it's a feature not a bug. Plus, I'm a huge fan of personalization.

I really appreciate you donating your time to my cause, though. You've improved my build, so thanks!

 

[artlessknave]

'm curious if anyone's done a benchmark test to show the difference

benchmark? the benchmark is reliability. this is known already; enterprise grade gear typically works so long it can be 3rd or 4th hand. desktop grade gear is known to be DOA.
reliability includes things like servers typically take 10-30 minutes to start up, checking functionality, ECC, all kinds of things. desktops barely check that they can boot at all.
its a case of you get what you paid for.