2

I have my RPI2 deployed as a receiver at remote location which is intended to be forever on. But after some 20-30 hours, it goes into some mode with just red light turned on(no green light blinking) and network shows host is down. The only solution is to manually turn off the power and turn it on. Is some solution possible so that my RPI stays on forever on and if it goes into that mode, it resets itself on its own without any Human interaction? (Please provide some simple solution as I am a newbie)

  • 2
    RPI have a watchdog timer, never tried it myself - blog.ricardoarturocabral.com/2013/01/… – Jaromanda X Nov 26 '15 at 4:09
  • @JaromandaX I read that blog on inbuilt watchdog timer earlier. It seems to be dependant on Raspban but the Raspban on Pi goes down once there is a power fluctuation. So do you think it will work? – Ric Nov 26 '15 at 6:06
  • Perhaps the Pi is just getting too hot? Is there enough airflow available? Allthough (some) people claim useless, those cheap stick-on passive cooling elements might give you just that extra cooling. – EDP Nov 26 '15 at 13:09
5

The conventional solution to this problem (as others have pointed out) is a Watchdog. There is a watchdog daemon, which you can install, but no one can design it for you, as this depends on what application you are running. Try reading its documentation.

You would probably be better to use a simple external timer (any solution which relies on software alone is inherently unreliable). This can be as simple as a 555 timer, which resets the Pi if not triggered regularly. I believe there are commercial modules available.

If you look through past issues of The MagPi you will find a case study (Project Curacao) which is instructive.

  • any solution which relies on software alone is inherently unreliable - the BCM chip the pi uses has on board watchdog - read the link I posted – Jaromanda X Nov 26 '15 at 4:46
  • @JaromandaX I am aware of this, and for most purposes this is probably OK, but it still relies on software. After a lifetime of running telecommunication facilities (many remote) I have learnt not to trust simple solutions. I used to tell management I could design a system to be as reliable as wanted, all they had to do was tell me how much I had to spend (which was usually too little). Telecommunication typically uses dual parallel systems (one active one hot standby) and they still crash. – Milliways Nov 26 '15 at 5:03
  • my understanding is that the software is used to prevent the watchdog circuit from reseting the chip - so, it's not reliant on software to detect a software issue, as such – Jaromanda X Nov 26 '15 at 5:07
  • @Milliways I have a arduino board connected to RPi using a logic level converter and I am planning to build my own watchdog sending Tx signals at some intervals. Any suggestions? Also can you tell me the reset switch for RPi? – Ric Nov 26 '15 at 5:59
  • @Ric The recent Pi have a Reset pin (labelled Run). Pull it to Gnd to reset. I would use a transistor or open collector driver. An Arduino seems overkill, but as I stated before I can't design it for you (particularly as you haven't bothered to tell anyone anything about you project. – Milliways Nov 26 '15 at 6:38
0

Designing and deploying a remote RPi mandates that there is a clear solid reliable communication channel available for managing it. Everything depends on how much free time is available to the RPi to perform self management.There are many simple solutions like clearing the logs, scheduled reboots, ensuring it can connect to the managing station and so on to reasonably keep its uptime high.

watchdog is a daemon meant to help in such situations. Its configuration allows several different types of watching itself. It does not imply the device will restart itself if it is completely down - it cannot unless it goes through the sequence as you rightly describe to get it back. watchdog helps before it reaches an unusable state

  • Designing and deploying a remote RPi mandates that there is a clear solid reliable communication channel available for managing it. - this stinks of a copy paste answer - the first paragraph is totally pointless - cite your sources – Jaromanda X Nov 26 '15 at 4:31
  • 1
    With respect, DOES NOT require "clear solid reliable communication channel available for managing it". It is normal to design remote loggers (I know of many installed in caves) where any form of communication is impossible. – Milliways Nov 26 '15 at 4:40
0

Overclocked maybe? Same thing happened to me, finetuned the overclocking a bit, now it's running fine. YMMV.

As for the watchdog, I've deployed that too. Tricky to get it working because the documentation/man page is not so easy to read. One thing that caught me for instance was that the documentation seems to assume the watchdog has a timeout of about 25 seconds, which the BCM does not. It has only 16 seconds.

An interval of 1 (the default) seems a bit quick, and is problematic when using using the ping test feature. Why? Because (ping-count*2) < interval needs to be true, otherwise the ping test will fail, your network interface is considered inop, and a watchdog timeout will reset the Pi.

These are the kind of details you'll have to work through when setting up the watchdog. It's smart to have a Pi on your workbench when trying out watchdog setups, and have a script ready to go on the Pi to 'keep the in its shed' on reboot, so to speak.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.