[Unit Review] Scale 25.10 on the YXK Iron Cow Zero1 Pro

TrueNAS General
,
21

Yes, for arm-based boards the partition layout is defined at a higher level. For whatever reason, they reused the chips meant for that in the x86 style boards too, which isn’t really a problem except for the fact that apparently we can’t delete the ghost partitions. But that’s not the actual annoying part.

They’re actually red herrings. The installer doesn’t actually care about them, it’s just that the system takes just a bit longer than expected for the partitions to reflect. I’m not exactly sure why the installer is so rush-rush on that, but that’s why simply increasing the spamming retry limit works: Eventually the partition layout does get updated and everything is fine, just that making the installer keep trying just a bit longer is what lets things go through as expected.

22

Yes, increasing the timeout does seem reasonable.

Please note the difference between “partition” and “device”. Since these are not partitions, but actual devices supplied by the eMMC, you can’t delete them without removing the eMMC. (I wish eMMCs were more programmable such that you could either remove them, or resize them.)

These “ghost devices” that eMMC has, act more like NVMe namespaces. Meaning the actual device does the “partitioning”, and not the OS. (Though in the NVMe’s case, a program can control NVMe namespaces…)

1 Like
23

I received my unit yesterday and had a similar experience. It failed on my first install attempt, but I pushed again and it seems to have installed from USB (final prompt said “The TrueNAS installation on mmcblk0 succeeded!”). Is there a SOP or best practice with which I can confirm that this did not leave any pre-installed software, etc.?

I’m looking for confirmation because I did not make any changes to the install script (as you later mention being a viable alt), my unit still flashes the “Iron Cow” logo on boot, and, upon booting, drops me into the GNU GRUB menu (which shows only TrueNAS Scale, Advanced options for TrueNAS Scale, and UEFI Firmware Settings, so seems fine?). My presumption is that this is as expected, but I’d appreciate, and feel more confident with, feedback

Today I will be working to catch up on some of the things we discussed previously, like the airflow having an unbroken path, seeing if I can swap in that Noctua fan, modifications to power settings in BIOS, etc.

That in mind, if anything has changed with those values or your testing since the 17th or so, don’t hesitate to mention it (I will certainly report back as well if I see anything of note with these values or similar)!

Also, any luck with finding an active cooler for CPU and/or testing largest single SODIMM it will accommodate? Thank you as well for the photos, and again for making this thread/the effort to investigate this unit and document the process.

P.s. If you want to see any of the install, etc. I went through, I have recordings of everything, as I have been doing a lot of this on stream, so that I have documentation for myself, and anyone else interested in this unit/general process – not trying to self-promote or anything, so provided only on request :blush:

24

Not really. The instant the mmcblk0 device gets erased it’s gone.

That’s just the burned-in BIOS image, nothing to note, it would require flashing a different BIOS file but it alone doesn’t really do anything.

Yep, this is exactly how TrueNAS works under the hood.

Haven’t really looked tbh. I won’t be mining on this and there’s a limit to how much you can really push the N100 CPU. I think being able to stably support 2.5GHz meets my needs.

As far as RAM, unfortunately due to the rampocalypse the largest sticks I have at the moment are 16Gb, and that works just fine. I don’t think it’s ECC so haven’t tried that.

IMG_20260627_195440626_HDR
IMG_20260627_195440626_HDR1920×2560 1010 KB

Top is the stick that came in the unit, bottom is the one I replaced it with. No issues so far. Theoretically it could do 32Gb, but I don’t have a stick of that size to try.

25

Thank you for all the confirmation. I’m not formally educated on this, and working mostly off things I’ve picked up by being adjacent to the space, so having clear verification really grounds me

If I may, I have a related question that is tangential to this topic that I’m hopeful you might be willing to consider, as it does relate to this enclosure

I had previously been using a TrueNAS instance on a mini-PC to learn a few things before committing to a NAS enclosure. When I ordered this, I pulled the media I had been using as a stand-in by exporting the pool from that instance. Since I was trying to find a cheap entry point to homelab build-out, I had been using this 2.5” SATA SSD in an enclosure to allow me to drop it into a USB-hub (externally powered), which was interfaced with the mini-PC. This set-up is not best practice by a mile, but was useful for my learning and never caused any issues

However, when I attempted to migrate this by exporting the pool, since I didn’t have a 2.5” to 3.5” adaptor for the NAS caddy, I just attached it via USB to the back of the NAS and imported it as a new pool there. I intended to migrate the data off that drive and remove it from the NAS entirely later that same day, but had some surprises that necessitated my involvement and time got away from me

When I returned to perform the migration, TrueNAS informed me that there were 3 errors with this drive, but I was not able to inspect them, etc. I spent time attempting to get more information as to the nature of the errors, interface with the drive via the TrueNAS shell, mount in on macOS/Windows, etc., but could not get it to respond. Windows recognized the Realtek chip inside the USB enclosure, but that’s it

As a Hail Mary, I pulled the drive from the enclosure and managed to get it to connect to the SATA connector of the NAS manually (i.e. “floating” without the caddy), but no dice

I can’t figure out a means of conclusively diagnosing the drive’s health, but I pulled the actual drive/PCB from the 2.5” enclosure and inspected the board, components, etc., directly – no signs of shorts, no burning smell, but also doesn’t seem to get warm when plugged in, etc.

Is it possible that this drive is still viable? Is this a common occurrence?

I’ve attempted to resolve this a few ways and tried to answer the questions on my own, but, as with the other self-taught skills, I’m never quite sure what qualifies as an exhaustive effort, best practice, or even what terms to search for that might unlock a solution to a problem that I had been calling by the wrong name, etc.

To wit, if there are any resources that you trust and/or general sources that are useful to you related to these topics, I’d be most grateful for the insight so that I may better myself and attempt to help others as well. I’ve mostly relied upon the docs for TrueNAS and resources like Lawrence Systems (though they aren’t equipped for the drive question above from my review)

But I digress

Thank you again for your help and insights

26

I bought one of these off eBay last week and have been fooling around with it. I put an name in it and setup UnRaid- relatively easy but it seemed not to like some of the flashdrives I had and refused to boot from them.

I setup both Ethernet using port aggregation, but am having difficulty even transferring at 1G to the nvme- I suspect the CPU needs some sort of active cooling- is there space in the case for this? Whole unit runs hot- have to jack up the fan just to keep it at 40C.

How about the stock fan? What size is it? I have some Noctua fans I’d like to replace it with.

27

There should be, but the heatsink provided doesn’t have mounting provisions for it. There’s about 2cm of clearance from the board to the drive chassis. See pictures up-thread for the type and hole arrangement.

Standard 120MM CPU PWM fan. I’ve replaced one in mine, haven’t done extensive testing personally between the stock one and one that has better flow but it’s definitely quieter.

I get about 2200Mbps sustained, but I think that’s a limitation of RaidZ2 with four 5400 rpm drives. Here’s some non-scientific stats from CrystalDiskMark over SMB:

image
image408×372 69.9 KB

Note that LAGG is a severely misinterpreted feature in the network space for consumers. It does not mean you can combine two 2.5Gb interfaces for one 5Gb interface in practice (i.e. if you have a 5Gb connection to the switch you put your LAGG-enabled NAS on, your PC will not actually see 5Gb performance). Please only do that if for some reason you have at least four clients routinely trying to pull the full link speed of a single interface and you know that the link is the actual bottleneck and know all the gotchas and pitfalls of such a setup. This user on Reddit summarized it well:

LACP is like adding more lanes to a highway, it doesnt change the speed limit, it just allows more cars to travel at the same time over the same road ALL at that same max speed.

LACP works OK for higher traffic, IE is you have many users hitting the same server, each user can get closer to 1GB each based on LACP count and user count but no one will get more than 1GB speeds.

If you want to increase the speed limit, you need to move up to multi GB links (2.5,5,10G)

This is a bigger up front investment but better in the long run as speeds for SSD’s and storage and everything else get faster, your 1GB links just feel slower as they are becoming the bottle neck.

Im running 1GB LACP to most of my NAS devices since i have the port capacity and i have multiple services requesting different data so my paths get used BUT i recently upgraded my network infra and now have 10G backbones for all switches into an aggr switch and a 10GB interface out to the WAN for my 5/2GB wan connection.

Does any one nas get more than 1GB speeds, NO, BUT, i can get 1GB speeds to EACH nas at the same time now since i actually have 10G backhaul.

28

The unit should have come with some small screws to do this, you would use the bottom holes on the sled.

Hard to tell, probably better as a new topic, though I would start by putting it back to the way it was before with your mini pc and see if it works that way. Y’all should be able to find the admonishment against a single-drive pool so I won’t rehash that here, but if you can get the other system working I’d probably focus more on getting your ”real” system working as you want it and then replicate the data over to is.

29

Thank you again for being willing to field this inquiry at all

The new instance had no issue importing the pool exported from the original TrueNAS instance (i.e. mini-PC). Consequently, once I saw that, I moved ahead with decommissioning the previous instance, so I don’t have the option of going back to exactly how things were before. However, I am not sure that would be a useful troubleshooting step, since the drive was recognized/imported by the new instance of TrueNAS, I verified the data on it, etc. It was only once I stepped away for a bit that I returned to the errors.

Y’all should be able to find the admonishment against a single-drive pool so I won’t rehash that here

Totally understood. The “real” instance I have set-up on the enclosure under consideration here is 2x pairs of mirrored drives. Also, for what it’s worth, the data on this (ostensibly) failed SSD was not singular/unique copies – I was just using this for learning purposes, so I never put myself into a position where a failure would be game over. I did, however, spend some time playing around with all of this, and, as part of that, cleaning up/deduplicating data, organizing things, etc.

So overall not a huge deal data-wise. I’m more concerned about not wasting time with duplicate effort (though not hard to script the flow of these files, etc., so not the worst problem to have). That said, the biggest concern I have is that I bought this SSD 1.25 years ago, and an identical drive has increased 300% since then, so it would be nice to know whether the drive is recoverable (whether with or without data is not important to me)

Anyway, I hear your advice related to starting a new topic, so I will look into doing that (though any thoughts you have related to the information here would also be very appreciated in the interim)

Thank you again for everything – I’m super grateful for the kind reception here

30

People should be aware that UnRaid is telling me the stock case fan is 3000RPM in case you intend on replacing it.

As far as port aggregation goes, with two 2.5G shouldn’t I at least get that much? My router/switch supports it (I think). At least I am using two of the 2.5G ports and (AFAIK) it’s configured properly. I suspect it’s something I’m doing in UnRaid as I’ve never used port aggregation before. I will try just a single 2.5G setup and see how it goes. Plus my previous server had a 10G card, so there’s that. It shouldn’t really matter though because I rarely ever transfer directly between two SSD/NVME drives anyways- basically just to 3.5” drives.

31

The stickied answer includes some Bios settings, etc., that you might find useful

Regarding aggregation, if you have infrastructure capable of providing 2.5G across the route you care about, I don’t see why not (unless kneecapped by CPU, read/write on the disks, or similar). That said, speeds are often lower than advertised, so in the best case 2.5G, and not likely faster

As Tsaukpaetra mentioned, aggregation is more like adding lanes to the highway. Speed limit doesn’t change, but number of drivers capable of achieving that limit does. They also mentioned earlier that this particular device’s CPU may falter under the pressure of maintaining 2.5G throughput on both ports (at least without some of the changes they made in Bios), but I believe they did ultimately see success with achieving greater than 1G throughput, if not managing to hold 2.5G

32

In theory that would be the max, so in theory you should get at best 2.5G. But better than 1G in any case. I’m definitely interested in what happens when you go back to a separated mode.

This. By default it limits the CPU to 6 watts after a 30 second “boost” at 10 watts, and I was definitely feeling the lower performance. Tweak that up according to the second post and that should definitely help things, before I fixed that it struggled to maintain 1G speeds in a speedtest (no disk IO).

33

I got my transfer speeds up by using an individual port and turning off IPv6, various plugins say I’m maxed out at 2.5G up/down - I’m writing around 127MBs to the array from a NVME on my PC- not very high initial boost speeds - could be going from a 10G based NAS to a 2.5G one. In any event this setup uses way less power than the old Lenovo Xeon based tower I was using with a 650w PSU so this is fine.

Next thing I need to do is fool around with the fan and bios settings and see if I can cool it down more and keep it quiet- stock fan is LOUD.

34

I got mine today. I was surprised to find that the emc on this is industrial grade according to Google, part SDINBDV4-64G. I’m experimenting with a 120mm low profile fan above the heatsink to cool the entire board. It just barely fits with some zip ties

35

This. It’s one of the first things that lead me to wonder why it felt like there was no airflow lmao.

Yeah I don’t believe the hallucination engine. It’s most likely just a standard eMMC chip actually, and the fact that it drops down to basically 300mbit speeds when imaging tells me it’s doing the “use a portion of the flash as SLC to make write speeds snappy” trick that all cheap SSDs do.

https://www.mouser.com/ProductDetail/SanDisk/SDINBDV4-64G

36

Industrial usually refers to a larger, usage temperature range. Sandisk listing on their website.

1 Like
37

Yeah I was wondering about the part number given, Google was adamant at correcting the DV4, which struck me as weird and probably a made-for-China thing.

38

Its not the fastest but spec sheet supports a good amount of p/e cycles, up to 3K. For something this cheap I’m happy with that

39

Yeah, better than cheap-o flash drives we used to use. :rofl:

Shouldn’t be too impactful since the boot pool should be primarily read-only anyways…

40

I would make the eMMc my boot drive but I figure I might want to sell it one day so I’ll just leave it alone. Unraid boots fast enough off a flashdrive anyways.

Going to see if I can zip-tie a low profile fan on the heatsink and definitely replace the stock fan. I’m thinking something could possibly be done with the pcie slot on the bottom, maybe with a right angle cable, or something.