You are welcome!

You are welcome!

:-)

I can't seem to find any of these tutorials that will work on a Pi 4. I fear I'm making things worse by following all of these various tutorials for a stupid shutdown switch and nothing works.

:-)

For rebooting you just have to switch power off and on...

You're welcome!

Thank you!

Glad to read that!

You're welcome!

Glad it helped!

You are welcome!

Thank you for sharing your experience!

I have SATA SSD. I could have NVMe SSD but that would be wasteful as the Pi woudn't benefit because ultimately whatever you do you meet USB 2 as far as I can see/ work out/ understand- and that appears for everything- everything to one USB 2. There are benefits to a SATA SSD, it completely sturates USB2 obviously, you'd really want USB3 which I have in a partial sense but I haven't created anny cache and for many files whatever cache I did create would not be enough. No the benefit is resilence compared to cards or sticks. Big storage at very cheap prices (120Gb is under£20, 250 is £30, 500 is £50) and it can be my boot drive. I haven't done it, or even know if it can be done but if the sMC could be used as a sort of scratch card that might be cool- what ever I think it should be able to be used as a card reader, which would be handy for copying ile from the Pi to give to someone else.

@@twmbarlwmstar I think it is old news, for example watch prof Christopher's video from 2 years ago: kzfaq.info/get/bejne/q8ieqtmpq5mvnJc.html I was talking about old ssd sold on ebay or possibly if you decide to upgrade your laptop ssd from 128GB to 1Tb as they are now more affordable than they were 2 years ago. for ssd vs sd card, have a look at linus tech tips video, kzfaq.info/get/bejne/acyioaWg1tetj4U.html ssd have a computer inside, sd cards are somewhat dumber. hence the difference in price. there are so many discussions on ssd vs sd card. For example, have a look at the review of the newly launched cheaper (but slower) qvo ssd from Samsung by anandtech: www.anandtech.com/show/13633/the-samsung-860-qvo-ssd-review and one of the popular nvme 970 evo vs 970 evo "plus" www.anandtech.com/show/13761/the-samsung-970-evo-plus-ssd-review

There are excellent tutorials on how to make the Raspberry Pi boot off a read-only card (e.g. hallard.me/raspberry-pi-read-only/ ). I tried that method in a heavily polluted industrial environment (e.g. tig welders, heavy motors aso) and I never had an SD card corruption, even though I had written a script to store data on a periodical basis. The trick is to enable write access only when actually writing to the card, to minimize corruption risks. Also, a GPIO can be used with an LED to indicate when the Pi writes to the SD card - although this is already done through the small, green board LED. All that matters is to have the *operating system* read-only, which Linux can do without an issue.

Glad it was helpful!

Thanks for the tip with "man". Resetting was already a good trick in the old Windows times ;-)

You are welcome!

Well possible...

This is what most wifi router vendors do too. Read only Squashfs and a small writeable storage area for settings.

Interesting idea. but is the problem not only shifted to the USB stick?

@@AndreasSpiess only for temporary data. USB storage is not used for booting, so I can handle that errors.

You're welcome!

Just curious: what is you ‘fleet’ used for?

@@Conservator. Hi Andrea's. Streaming server for LiveATC another for LoRaWAN on TTN (thanks for those video's), web server and node for portable FT8.

You are welcome!

Darrin Pearce Impressiv! 👍

You are welcome!

Good to know. I did not know that and I do not know why they used this pin then. At least for safe shutdown, I showed a second possibility. And maybe with an open collector transistor in parallel, you could even restart it without harming i2c

me too

​@@AndreasSpiess Could you please show a small schematic for the pin 3 workaround with a transitor?

:-(

You are welcome!

I hope you liked it...

@@AndreasSpiess yeah of course, the amazing landscape

Yes, but maybe mechanically less elegant. For me, it was not necessary as I usually know what I do ;-) (I grew up when mistakes destroyed things and we had to think twice).

You are welcome!

Thank you for your hints. Since I always have plenty of SD space I did not care about the (too big) swap size. Do I have to fear anything?

Hm, depending on the number of processes and their respective memory needs your swap space can grow - in theory - without bounds. In all practicality there‘s most probably nothing to fear, but for „production“ systems it is a conservative approach to nail down everything that might move. Ok, let me reformulate: For systems you use for experimenting, including e.g. tensorflow and other memory-costly apps, your approach might be the best. But if you have something that runs a well-defined number of applications in your home automation system installed somewhere you can‘t reach easily, you should use MAX_SWAP after you have watched the memory consumption of all your apps for some days. Calculate the maximum memory needed and use a factor of 2 or 3. That would be my approach, but ymmv.

You are welcome!

You are right.

Maybe an old Firmware/operating system (This feature seems to be quite new)

Watch my earlier videos. There, both run on a pi zero. But it is not an ideal solution. If you can afford, go for a bigger one.

You are welcome!

You are welcome!

Thank you!

I am no SBC guy and because I am a Linux noob I stick with the Raspberry because of its community

It is the same as if you type „shutdown „

I never did something like that. I trust that it is shutdown automatically (gets a signal from the OS). But I do not know. So far I never had problems.

I thought I mentioned the reason (connector is flush to the end). And I also mentioned the rebooting.

:-)

Good idea!

For the moment it is not available. And I have to wait till it is supported by the Arduino IDE

The shutdown has to start before the power goes down ;-)

Most power banks are not useable as USV/UPS, because they do not allow simultaneous charging and discharging.

yes i know! But i use one how can charge and running my pi at the same time. Some pi is running with the poe heat and powered over a switch with a ups behind.

I assume because the engineers at the Pi foundation or at Broadcom decided to be like that

You can get help on the overlay using the "dtoverlay -h gpio-shutdown" command. This also mentions the power on function: "This overlay only handles shutdown. After shutdown, the system can be powered up again by driving GPIO3 low. The default configuration uses GPIO3 with a pullup, so if you connect a button between GPIO3 and GND (pin 5 and 6 on the 40-pin header), you get a shutdown and power-up button."

@@silasparker So that restart works only after shutdown, I read that part in the documentation and assumed both functions are not enabled at the same time. (Force restart while shutting down would increase the danger of corruption, but makes sense its only active once the system is idle after shutdown.) Thank you!

They were both enabled in my case. I push the button once and it shuts down. After shutdown, I push it again and it restarts.

I searched and found some discussion on Stackoverflow on this: raspberrypi.stackexchange.com/questions/77905/raspberry-pi-3-model-b-dtoverlay-gpio-shutdown As mentioned by others, the behaviour is dependent on GPIOs setup wrt being default high or low.

Good idea...

I do not know the reason

Thank you for your information. So far I had no issues with the writing limitations but read about it. I usually use at least ram2disk to reduce some wear.

Some sd cards write when not in use, to move data to unused sectors as wear leveling. So you aren't completely safe.

@Alejandro Vargas: This is incorrect. The Pi is not corrupting the filesystem on the sdcard but it is physically destroying the sdcard. Sometimes you can be lucky and continue to use the sdcard but it is definitely going to break down sometime soon after this event. I had already like +10 sdcards of different manufacturers dying on me by exactly this Pi character trait. The underlying problem seems to be a physical write access when the voltage on the sdcard Vcc pin has already broken down or browned out. It also happens when spurious voltages of too high an amount is shortly present in the right (or bad) moment when power is lost. BTW, most of the corrupted sdcards still have a filesystem on it that can be used read-only. Sometimes the FS is completely done, but often you can at leas extract your data even if the sdcard is toast, write-access-wise. It is of particular nastiness if you got a Pi with Linux having panic'd. There's no way to shutdown via software. I ignore so far whether the dtoverlay=gpio3... trick would be a feasable solution in more cases than the one we're left with: unplug the power connector. I don't expect it to work in a case of kernel panic, there's just nothing left running to exec one more call to power off cleanly (especially regarding sdio shutdown).

Show me the system where "the boot can never be broken". Such a thing just does not exist. Particularly, when dealing with DOS partitions, as I understand you, you have to deal with vfat and consorts. Breaking a FAT is ****cking happening on it's own!

@@AndreasSpiess How is "ram2disk" reducing flash cell wear?

I never checked it with the Pi4

Raspberries do not switch on again after shutdown. You have to remove power first and then they boot

I to not think you missed anything. But if you switch the power off, you have no time for that.

I do not know as I do not use noobs.

Seems to be a huge system ;-)

I looked at it and I am not sure. You do not need debouncing of the GPIO3 pin, and the power has to be cut off by the RPI after safe shutdown, not by an external IC, I think.

@@AndreasSpiess The bush button control is debounced, so no false triggering. The LTC2954 has an enable pin that is used for controlling the RPI's psu either on/off. Starting the RPi up, the press of the pushbutton re-enables the RPI's psu and this starts up the RPi. On shutdown, arrange for the KILL pin to be set low and the LTC2954 will disable the RPI's psu. Setting this pin has to be done as late as possible. I managed to do this by creating a delay from when the RPi o/s signals a terminate and then set the KILL pin low after about 10 seconds. This could be adjusted to a longer/shorter delay if required. The crucial component is having a PSU that has an enable/disable facility, I used an LM2596, which has enable/disable and provides 5v at up to 3 amps. I think the LTC2954 is a very usable device as it has a very efficient 6uA supply current so is ideal for battery use. The ability to switch off the RPI's psu is great as both the RPI and its psu are both effectively switched off. I believe the LM2596 will only consume around 80uA when in standby.

You can buy such devices if you want. I also made videos concerning this topic.

I think it will be sure. But mechanically a little more challenging...

@@AndreasSpiess You can use a SMD resistor even below the button. Or a regular one if you partially breaks the top plastic part of the connector to fit it in.

Gern geschehen

If the Pi is without power and gets power it automatically starts. If you want to start a powered Pi you can do that as shown in hte video. And if you want to do that remotely you have to find a way to toggle this GPIO from remote. This can be done in many ways, for example with an ESP8266 or a 433 MHz button.

@@AndreasSpiess Thanks Andreas. One then shifts the problem to the ESP8266. How does the ESP8266 know that the UPS has lost power & the Pi has powered down (easy bit: the Pi NUT send a MQTT message to the ESP8266 before shutting down). Next the ESP8266 needs to know the UPS has not shut down as power has been restored so the ESP8266 can toggle the GPIO. Maybe the USP can tell the ESP8266 that power has been restored and then the ESP8266 can toggle the GPIO - but how can the UPS communicate this to the ESP8266? However, if the UPS has shut down (prolonged outage) then hasn't the ESP8266 also shut down? Maybe the ESP8266 is battery powered. Or maybe the ESP8266 has a start up routine when power is restored that checks if the PI had previously shut down (record written somewhere). Maybe the ESP8266 interrogates the Pi and if the Pi responds it knows the Pi has auto restarted after a prolonged power outage so there is no need to toggle the GPIO. I'm thinking aloud here but I think the key thing is that the logic has to be bullet proof for all scenarios.

An ESP has input pins to measure voltage or digital signals.

I did not look into that topic so far.

Thank you for the feedback!

I think it should be open, but I never tested what happens if closed during power-up.

@@AndreasSpiess Thanks - I did also find a few other sources that suggested a normally-open, momentary switch--apparently a normally-close switch can be used as well but the code would be a bit different.

Bringamosa 2times yes

I had issues but was not able to find the root cause. If you search the internet you for sure were lucky!

Most filesystems try to avoid unsafe states at all time. Some write new changes as planned changes into a journal, execute the changes on the real data and only if these changes are completed, they are removed from the journal. If a crash/poweroff happens in between, then it will just re-apply the changes that were written to the journal or it doesn't do anything to the data. This allows to always have a well defined state of the filesystem. This is called a journaling filessystem. Example: ext3/ext4/xfs Another variant is, if data is modified, to write a complete new data structure that references parts of the old data structure. If you change one byte of a large file, the following happens: It copies the block that contains the modified byte and applies the modification. Then it copies the data-structure that points to the many blocks of the large file, but it replaces the pointer to the modified block. And in a final atomic write operation, the new data-structure is written to some other data structure. This is called copy on write filesystem. Example: btrfs/zfs Both methods have some drawbacks. Journaling filesystems need to write data twice. Copy on write filesystems tend to be CPU-intensive when randomly accessing data in large files (Databases). So people sacrificed data integrity to speed. The typical solution is, to disable journaling or copy on write for data. These techniques are still used for meta-data, so your filesystem structure stays intact. File content can still be messed up. On ext4 this is the default, every write to data goes directly into the file. It is the responsibility of the program that writes files to do this safely. Data-journaling can be enabled with the mount option "data=journal". On btrfs, the default is data-integrity. Everything goes through the copy-on-write mechanism. But it is possible to set the 'C' attribute with chattr/lsattr to some specific directory prior to creating files in it, this disables the copy-on-write mechanism for the files in the directory. Typical directories für this are /var/lib/libvirt/images/ or /var/lib/mysql/. On ext4 you could mount / with data=journal and have a separate filesystem for the database that uses the faster defaults. Anyways, the filesystem metadata is relatively safe and data corruption is not directly visible. Programs like mysql can often repair their data files, but it is better to avoid these situations. Also, this has nothing to do with the SDCard. The SDCard can also be corrupted, some SDCards used unsafe methods for wear leveling (redirection of addresses depending on write cycles) and corrupted their own database where some block was stored. After power-loss they were unusable. I never had such a SDCard but I have read about these problems. If something like that happens, no filesystem wizardry will help you to recover your data. ;)

I have several pi's that have been shut down by power off regularly for years, no problems. But, I have also built others with an overlay and also no issues pulling the plug.

Thanks for the response. This button is so easy to be made it is almost stupid if i don't and risk the card to get corrupted

It can corrupt your SD card.

@@AndreasSpiess what about my Rusberry pi will it damage?

Usually not. If your power supply does not produce spikes when power comes back.

As usual, there is a link in the video description to the commands. It is the same as above.

Corrected. Thank you!

I do not know. So far, I never used I2C on a Raspberry. I use MCUs for my sensors.

I once made a video about this (without the gpio solution, but with super caps)

@@AndreasSpiess I know. I think super caps at the input could simplify the system more especially with gpio shutdown. No need for lithium chargers or battery monitoring circuits. Perhaps a diode and a resistor would be enough to sense power loss.

I was no more able to boot the Pi with pressing the button. With GPIO3 you should be able to boot the Pi without re-powering it.

@@AndreasSpiess Ah I see, with 3 pushing the button can turn it back on? That’s extremely handy

If I remember right: Only the Pi4. But I am not sure.

Thank you for the tip. I did not measure the current when the RPI was off

Unfortunately I cannot perform remote debugging. This behavior of your RPI should not happen.

I have to admit I am a Linux noob. Maybe somebody else knows it?

I use Motioneye OS and for my purposes it runs very well on a Zero W. The author has specific images for the different versions of Pi.

Maybe you watch the last video. There we try to build a UPS with such a function. And you can use a switch to GPOI3 to properly shutdown the Pi4

@@AndreasSpiess Ok tx Andreas I am going to watch it!

Yes

@@AndreasSpiess 😅 I try to be attentive. 😉

I do not know.

as far as i googled, the "dtoverlay=gpio-shutdown" uses gpio-keys driver to trigger a "power-key event", explanation here: www.stderr.nl/Blog/Hardware/RaspberryPi/PowerButton.html i think you can drive it further an map all gpios (github.com/notro/fbtft-spindle/wiki/GPIO-keyboard) to normal keyboard keys and trigger events with triggerhappy (raspberrypi.stackexchange.com/questions/15337/gpio-button-to-control-backlight-on-pitft) . at the end, there must be a way to do any script, i.e. reboot instead shutdown ... @andreas: thanks for inspiration :-)

As said: You can use it to shutdown, But it does not boot anymore.

@@AndreasSpiess Roger on that. Thnx

You see it if the RPI LED stops to blink

Good tip!

Maybe you watch my other videos about this topic. There you even find supercaps, if I remember right ;-)

If you search for the links: They are in the comment below each video

libreelec.tv/ says: Yes.

Strange. Do you use GPIO3? Maybe they changed something in the OS.

Nothing for me :-(

I am not sure. The real problem is pin 1 and 2.

@@AndreasSpiess I'm referring to using GPIO2 set to 0v output in place of the ground pin. You'd still use GPIO3 the same way so you'd be using pins 3,5 instead of 5,6. If accidentally reversed, the switch would connect pins 2,4 which are both at the same 5v potential so no problem. Pin 1 (reverses pin 6) no longer comes into the equation, so again, no problem.

It will only reboot if you use GPIO3. All other pins only shutdown

Strange. ssh without extension works everywhere else.

I think I made a video about that.

@@AndreasSpiess Do you remember which?

No

@@AndreasSpiess Can you research and let everyone know? Would be helpful, thanks!

So far I did not experience those issues (I add the files on my PC)

The Pi has no way of doing that. You need external circuitry to have that functionality.

@@zaprodk I thought so then its useless to me at least I need an indication for powering it off.

@@siskodata Google "ATXRaspi" - Yeah, ATX is completely wrong in this context as it has nothing to do with the ATX configuration, but someone made a module that does what you want.

@@zaprodk : lowpowerlab.com/guide/atxraspi/ lowpowerlab has some interesing things to give, andreas did present something from there before, i guess...

You can notice by seeing that the Ethernet LEDs, if it's connected, and the green led next to the red power one are not blinking after a few seconds

:-)

Good idea!

Thank you!

Thank you for your information!

I think that ic was there on AliExpress. I once killed the ic and looked for it on the internet. luckily it was loose solder so did not have to buy that.

@@dd0356 The chip is on Aliexpress. Problem is that it's not the correct one.

It’s not the standard version of the MXL7704 chip even tough the label doesn’t indicate it. Just google it. You’ll find plenty of stories about it...

Hackaday, 4 days ago: hackaday.com/2019/06/12/shorting-pins-on-a-raspberry-pi-is-a-bad-idea-pmic-failures-under-investigation/#more-355336

:-)