Raspi Green-NAS (Serial and Relais HAT)

I needed some extensions for my homeserver consisting of a Raspberry Pi4, so I finally decided to build an extension board on my own.

I'll try to avoid the term HAT in the following instead I will be talking of an extension board because that's what it is.

I'm not a friend of people that keep on inventing new terms for old socks. – HAT, Shield, Sketch — App...

Function

Here the completed board, plugged on my Pi4B.

The board contains four coil drivers for connecting two bipolar relays.

Additionally, UART5 is layed out as a RS-232 interface via a 9 pin Sub-D connector (like on a PC).

In general, the coil drivers should work with any Pi having a 40 pin connector, i.e. starting from Pi1B+, the serial interface requires at least a Pi4B.

An anecdote by the way: the resistors and diodes you see in the photo are actually from my assortment that I purchased in my apprenticeship, about 45 years ago... which were still solderable without problems, without storage under inert gas or the like! At that time, metal film resistors were not yet very common, respectively expensive and 5% tolerance was the state of the art, which is still more than sufficient here.
In fact, I still remember resistors with a silver tolerance ring (±10%) or without the ring at all (±20%)

Circuit and layout has again be done using KiCad. Meanwhile I must say that KiCad proved as a really professional tool for creating schematics and PCB layouts. It surely has some disadvantages compared to competitors but it also has powerful tools and if you work with it for a while (and did not rely too much on the provided library) you know what can be done in KiCad and what you'd better import from LibreCad you can work accurately and fast. Additionally, the possibility to directly edit the human readable files proved very time saving in some cases. A simple search-and-replace can do half of an hour work in a 20 k$-suite of a competitor.

I mounted the relays into a plug adapter (Goobay 68906). This was available from my distributor for only €1.70, what is a tenth of the money they want for an empty case. An old telephone cable serves as connection with a perfectly fitting twist-proof strain relief.

I dismounted the switch and replaced it by the relay. It fitted perfectly into the space of the switch. It has nearly the same dimensions and covers the hole of the switch so there is no danger that anything could get into the case producing a short or a dangerous situation.

Well known NAS servers consume an unnecessary amount of power, especially if you don't use them permanently. I used a Thecus N5550 for some time, consuming 26 watts of power even in standby, that is about five euros per month only to be there (active: 50 W). For €60 a year I could go out for a really good dinner on New Year's Eve. We can do it cheaper!

Indeed, using multiple disks in a raid array for most of the private users is simply overkill. A regular backup is much more important. Raid can not replace this. You can get 3.5" disks up to 26 TB now, what should be more than enough for any private application.

The majority of private and Soho administrators would probably only notice the failure of one drive when a second drive also fails and access is no longer possible. Sure, it will last longer, the system will work longer, but a failure is a failure! Who will pay attention to whether something is flashing or glowing red on the NAS? Hopefully they at least have a backup then!

If some crypto-trojan encrypts your disk, this will happen relatively fast (well, hours but you would not take notice of it immediately) and even your raid would be encrypted over night!

RAID can not replace a Backup!

Therefore, the backup disk must be regularly replaced by a second (third, fourth...) so that there is always a disk available that is not online and would therefore also have been encrypted!

A backup can only be replaced by a more frequently backup, so frequent that a ransom demand is not profitable for the attacker or you can say a restore is cheaper!

Hard drives are cheap. Cheap compared to the potential loss of data, even in private life (think about it, all the photos of your life are gone... what would they have been worth to you? Now or in 30 years?). Buying a new one every year and keeping twenty of them in rotation is nothing compared to the potential loss!

As a home user, you may not need twenty hard drives, but each additional one increases the chance that you can recover something irretrievable, even if you accidentally deleted it years ago and only now realize it.

Remember to mount the backup disk read-only in this case. This way, a virus that may still be active cannot make it unusable (unless it is intelligent enough and found a way to mount it somewhere else R⁠/⁠W!) An NFS mount from a definitely uninfected system might be the solution, or a copy on a USB stick, which could be replaced if necessary. In any case, be careful with the two or three backup disks! Use only copies of them, that you have made on a clean system, even if this costs extra time!

The complete loss of the live system must only cause relatively low costs. For me as a private user, a weekly backup is usually sufficient, for a medium-sized company perhaps daily, for larger companies rather more often (or more differentiated, current data must be backed up more oftenly, archive data rather just every now and then).

That's why I not only have backups of the software, I also have a second Pi4B in my drawer, with a slightly older but working SD card which can replace a failed system within half an hour.

If you really need more than one disk for NAS, this would be the time to consider a dual docking station or a Mobius case as well as software raid. I haven't done that.

Sounds fast but it can take its time. From one docking station to the other my Pi4B makes a terabyte in about four hours. A full backup of 6 TB would take about a day. If the target is not new and you only have to sync the difference to the last backup, it would be much faster, of course (in my system, hosting an 8 TB disk, normally in about 15 minutes).

My backup script is additionally designed in such a way that the backup disk must of course be formatted, but must also already contain each (root) directory to be backed up, at least as an empty directory. So you can use e.g. smaller disks as a backup target to backup only certain or frequently changed directories.

Unfortunately, this feature does not exist for unmount! Here comes my way:

The mount script turns on the disk, waits 20 seconds until it is ready and sets an appropriate sleep time via hdparm. So the disk at least goes into standby if it is no longer in use.

One problem is that Autofs seems to cache the result of the script for some time. So it is not always executed! Obviously Autofs is not made for a task like this...

This is, of course, not optimal but it comes very close to the desired behavior.

A Sharkoon dual dockingstation failed because it does not have a power switch. It must be turned on by pressing a button every time you power it on 😠. Why should anyone build such a crap?

In my experience, nothing at all. My Pi4B has 1 GB of RAM. It works perfectly and fast, without any notable penalty.

More RAM only costs money and rises energy consumption. I would recommend to use the smallest Version available, it will be absolutely sufficient.

The Pi4B with 1 GB of RAM actually (end of 2024) is at about 40 Euro, the one with 8 GB at about twice the price, without any noteable advantage in our case.

The costs are about €6.50 per year for the Raspi, even if your disk would be permanently spinning this would not be more than €40 per year.

To give a comparison: using the professional NAS with five disks mounted you'd have to pay €57 just for stand by and €110 for the permanently active system. Since especially in a home environment the disks are used only rarely, we can talk of a saving by a factor of 6..8.

Of course, the professional System has its advantages which it may play out in an environment with tens of users working day and night, it would be much faster then. In addition, the failure of a single hard disk could be repaired by an attentive (!) administrator without interrupting operations or losing data. But for most of the home and SOHO users, this has no real meaning.