Kobol 64 NAS Data out

Hardware Hub
1

https://wiki.kobol.io/helios64/intro/

@SgtAwesomesauce @wendell @Eden (if I left anybody out that I recalled was slightly interested in the device im sorry)… data and open-ness looks encouraging

https://kobol.io/

The helios64, Just released. Good first spec revisions and tbch I think it might just live up to a bit of its hype. Dont take what I say below as knocking at it. This is me nit picking. It looks like a great device and once I see some positive solid reviews I will probably purchase it

Alright so some engineering issues I have that I am going to nit pick at which by all means could be improved im sure with firmware and other upgrades are the following

The battery pack is rated 7.2V 3180mAh (22.9Wh) with max continuous discharge of 10A. Thats cool and all but my issue with it is looking at powering any HDD’s with it would result in a very short discharge time. Now one could make a bigger battery pack but they note that they are using the TI BQ24133 charger & power path IC which does a good job at charging the battery in three phases: preconditioning, constant current, and constant voltage . Here is the thing it cant do. You cant limit discharge of a bigger pack between 30 and 80 percent. I suppose thats fine but not good for 18650 longevity. From the SoC / OS you will only be able to get the charge state. I wish it could release more data but Im pretty sure thats in the firmware and can only be addressed via serial… Again im sure one could hack the firmware a bit but thats not recommended.

Now from what I can gather the TI also has an unfortunate built in charging time limitation. It will stop charging regardless if the cells are not full by 9 hours. The 2 serial pack takes 4 hours to get to 100 percent. So a bigger battery pack would likely be limited to a 2S2P configuration which means at max you could double the capacity but in no way could you increase the maximum current discharge. So the batteries wont power disks. It will probably tells the system to shut down the disks and just power the ARM part of the NAS. At least I hope they do that so the UPS can be on for longer. It would defeat the point of the UPS providing full functionality. IDK 15 minutes full disk run time just seems a little weak

Now looking at their board design its possible you could change the resistor and allow for a bigger/greater current discharge and according to the TI data sheet you could change the resister on it to up the 1A charge current to 2.5A. IDK what this would do.

Its cool though I like what I am seeing. I think theres definitely room for improvement. UPS design wise they could do better. Overall Its pretty damn solid. I think the flashing process is a bit wonky but thats just ARM for ya

Id be interested to see this… I know its not standard but M.2 Wifi AX card. Be able to just store it away in a corner without ethernet. The latency is low. Wifi AX is reliable and its incredibly fast for what it is. It would allow for flexibility of placement. I just dont know if it can be done because its a SATA 3.0 connected M.2 not PCIE. Because of that im pretty sure only bluetooth would work because the system interface specification for intel wifi ax 200 is Wi-Fi(PCIe), BT(USB). So I guess the only way this would be possible is to wait for someone to release a USB 3.0 USB Qualcomm based Wifi AX module. Might take time but AX is kind of a generational step where I see more things that were wired becoming wireless which is something I really desire. Wires are lame lol

3 Likes
2

Mine is on a boat. Will update when it arrives.

2 Likes
3

So with the weak battery, would it work better to bring the machine to a graceful shutdown (cleanly writing out to disks and recording power loss etc) , rather than to keep operating during power outages?

Or is the NAS supposed to run for a few minutes, in an anticipated short duration black/brownout?
And if the battery was to keep the unit operating normally in a blackout, the switch it is ingesting data from/serving to would have to be separately UPS’d.
In that scenario, why would someone have the rest of their equipment covered, but not the NAS?

I guess I’m just asking if the weak battery might be just enough for the write out, or more likely saving state like a raid card’s battery?

2 Likes
4

Personally that is what I would do. I would write a script to do so IMHO. Data integrity wise that would make the most sense because over time those cells will become unpredictable

They estimate at least a 20 minute run time. Im assuming its to keep alive during short power outages… Im not sure

Probably so. Not sure. the product has yet to be fully released

@SgtAwesomesauce I would love to see some on L1 or L1 themselves review it. I think its a cool little development for the Open Source work :wink:

3 Likes
5

Thanks, looks like a nice little machine, all in all

6

@PhaseLockedLoop - Do you want me to hijack your thread, or start my own?

https://forum.level1techs.com/t/post-what-new-thing-you-acquired-recently/149881/4545?u=imhigh.today

1 Like
7

I think I have to agree with the whole needing just one pci-e slot for ad in wifi /bt. It would make it super handy for someone like me.

8

hijack

Go for it. LOVE THIS… DO ITTT!

1 Like
9

it doesnt serve via actual PCIE. Its a SATA. So you cant really use an AX wifi card so yeah that sucks

10

This appears to be true Linux-side using the Armbian 5.8.13-rockchip64 kernel, at least for now.

/sys/class/power_supply/gpio-charger/* output gives

POWER_SUPPLY_NAME=gpio-charger
POWER_SUPPLY_TYPE=Mains
POWER_SUPPLY_ONLINE=1
POWER_SUPPLY_STATUS=Charging

The latest Software Update blog post from Kobol mentions that power level monitoring and control is not yet functional due to a lack of userspace tooling to monitor power level and initiate shutdown.

That tells me two things:

  • Their intention is that the UPS function as a buffer to allow for safe shutdown, and
  • They believe that discharge state will be exposed to userspace eventually, implying it’s exposed somewhere (just not in mainline linux yet).

So the batteries wont power disks. It will probably tells the system to shut down the disks and just power the ARM part of the NAS.

I’ve tested this using 3x Western Digital SATA spinning platter drives. The entire systems remains fully powered when the DC cable is yanked out of the back of the box. This includes the disks, LEDs, and NICs.

I sat and watched the output of dmesg, lsblk, and the sysfs power state. The only change is that charging status changes from “Not charging” to “Charging”.

I tested actual reads and writes to rule out phantom devices being visible and things work normally.

2 Likes
11

I know you were talking specifically about the UPS device, but this statement is very true.

Before I start listing all of the “problems” I see with this device so far, I want to be clear that I’m very much the target audience for this device and I’m happy to have it.

If you did not notice the word kit on both the board and the enclosure, it’s important. This is a NAS kit, not an off-the-shelf consumer product.

I made a mistake in the “What did you acquire recently?” thread by calling some of these parts “accessories” when they’re not. The cable ties aren’t perks to aid with cable management. :laughing:

IMG_20201006_164710
IMG_20201006_1647104160×3120 8.07 MB

IMG_20201006_190708
IMG_20201006_1907082340×3120 4.38 MB

It’s very, very obvious to me that they kept costs down by decentraling the labor costs of running the wires.

IMG_20201006_164830
IMG_20201006_1648303120×2340 3.77 MB

And then there is this gnarly thing. It reminds me unpleasantly of both fire hazards and ketchup.

As I was securing the third SATA connector and tugging on the bajillionth zip tie, a specific scene came to mind…

This is probably my single biggest “complaint” about the kit. At it’s heart, it’s a small form factor build and those are notoriously frustrating. If Kobol were to offer a fully-assembled kit, the labor costs would be substantial and it would probably make this kit more expensive than comparable Synology or QNAP barebones kits.

I suspect that in future revisions, they’ll find it’s more cost effective to develop a drop-in backplane that simplifies the assembly process rather than reduce costs by shifting the labor.

And with that in mind, they’re actually in a good position to do exactly that. The inner tray is swapable, so they don’t need to redesign from the ground up to do that. They just need to design a better insert (and the backplane, of course).

It would be awesome if such a backplane were built around the System76 Thelio IO daughterboard. Convergence in the libre space can’t be a bad thing.

4 Likes
12

IMG_20201006_130702
IMG_20201006_1307023120×2340 3.92 MB

There’s some corrosion on my enclosure kit where the screw holes were drilled. I’m not terribly concerned, but I feel like a dab of grease after machining wouldn’t add that much to the cost.

I doubt “put it on the ocean for a month” was included in their factory acceptance testing. :stuck_out_tongue:

1 Like
13

How are you liking it so far? I was thinking about getting one.

1 Like
14

So far I’m only using it as a dead-simple RAID array using mdadm and connecting to it over ssh.

Using my HTPC to play video files has been the bulk of what it gets used for right now, and that works well. Youtube-dl works rather well fetching directly to the box.

I have a faster UHS-I SD card, and some new Seagate Ironwolf drives on the way. Once I’ve got my arrays in an acceptable final form, I’ll do some actual benchmarking, but I wasn’t really prepared for this to arrive. I still need to upgrade my network before I see how fast it can really go.

Armbian’s “Softy” installer is nice and convenient if you want to install stuff like Nextcloud, but this still requires as much command line experience as a Raspberry Pi.

It should also be noted (see earlier thread posts) that certain things like battery discharge state and auto-shutdown don’t work yet.

If you expect a polished consumer product, buy a Synology instead. If you want a fun DIY project or are buying primarily because this is open hardware, it’s good and should get better over time.

1 Like
15

I wrote “I bought a faster UHS-I” card, but only just realized… I’m running off of an SD card.

That’s remarkable. In terms of perceived responsiveness, I think this might be the most performant SD-based device I’ve ever used. My Pinephone and Raspberry Pis have never handled IO quite as well as this thing.

I also see that Armbian creates zram partitions for /var/log and for swap. Perhaps that’s why it feel more responsive. I shall investigate with my Pinephone… :thinking:

16

sd you say ?

17

MicroSD for the OS image, yes.

It runs u-boot, so you could configure it to boot from USB, pxe, or off the M.2 sata drive. But I like the simplicity of an OS on microSD and other mount points on $other_storage.

18

I thought any sata drive could? Its all the same bus?

19

So long as u-boot can read it, you can boot from it. Before you’ve loaded init, though, you’re not dealing with RAID arrays but raw filesystems.

I figured the only meaningful boot from SATA would be M.2. SD would be faster than spinning up a rusty platter.

1 Like
20

Cant uboot read the ext4fs from a RAID 6 Array?