SSH access and web interface stop functioning after a few hours

[vahegan]

Hi,

Starting this week, I am facing a weird issue: I loose web access to the web interface and cannot access the Pi via SSH after a few hours of work.

I first thought that the power supply might be the culprit (although I am using quite powerful 12V, 36W power brick) on a Robo-tank. But then I realized that only the IP interface stops functioning, whereas the controller continues to work - judging by ATO, at least.

When I power cycle the Pi, it reappears on the router list. The router is a powerful XiaoMi 3600, which is designed for IoT and can handle 256 simultaneous connections, so I don't think that the router is to blame - there are many other connected devices on WiFi and I am not facing issues with any.

Also, the system was not showing this earlier. I am seeing the issue since a few days only.

What may be causing this strange behavior? What can be done to debug and try to find out what triggers this connectivity loss?

 

[Sral]

Power sounds unlikely, but the PI obviously utilizes 5V and 3.3V, so I wouldn‘t rule it out completely. Found this interesting thing to read core voltage, but that might not give you enough info.

just a few ideas:

• Create a Cronjob to write values with timestamps into a file, or check reefPI‘s database/WebInterface if values are continuously written during the IP out
• Check the error log
  • Both in Reef-Pi web interface
  • And the ssh interface like This
  • Not sure if both are identical ^^
• What are you using for the connection ?
  • Does the PI have some kind off power save for that ?
• Personally, I have the PI running at a fixed IP over Ethernet, if yours runs on dhcp, maybe there are regular changes in IP ?
• I did have the problem in the past that my router updated and forgot a few settings, like my white list. Possibly because it wasn’t the highest quality, but you might check that. Also try rebooting your Router, that did fix a WiFi issue I had recently.

 

[robsworld78]

Hi,

Starting this week, I am facing a weird issue: I loose web access to the web interface and cannot access the Pi via SSH after a few hours of work.

I first thought that the power supply might be the culprit (although I am using quite powerful 12V, 36W power brick) on a Robo-tank. But then I realized that only the IP interface stops functioning, whereas the controller continues to work - judging by ATO, at least.

When I power cycle the Pi, it reappears on the router list. The router is a powerful XiaoMi 3600, which is designed for IoT and can handle 256 simultaneous connections, so I don't think that the router is to blame - there are many other connected devices on WiFi and I am not facing issues with any.

Also, the system was not showing this earlier. I am seeing the issue since a few days only.

What may be causing this strange behavior? What can be done to debug and try to find out what triggers this connectivity loss?

If you have another power supply it might be worth trying, seen a few strange network related issues from weak power supplies. Any chance you moved the controller last week?

 

[vahegan]

Thanks for the responses. I am on pi Zero2, so no wired connection, only 2.4G. Yes, I tried rebooting the router, no difference. Then I was power cycling the controller yesterday and still it wasn't coming online: the router wasn't seeing it. I disconnected everything and brought it very close to the router, and once I was able to see that it got an IP (the DHCP on the router always gives the same IP address to that MAC) I connected to the web interface to download the database. Once I did this, I took out the SD card and reinstalled raspbian and reef pi. The other thing I changed is I connected to 60W, 12V MeanWell power supply that I had purchased specifically for powering Reef Pi and connected devices. I then re-assembled everything in place and uploaded the database last night. It seems to be up and running since then - almost 24 hours now and no issues so far. Don't know what did the trick: the updated image, or the good power supply.

 

[Ranjib]

as others called out , we know wifi drop out can originate from either power or router issue. If it was a power issue, reef-pi should love errors indicating system under power, if it was a router issue i would expect other things also to show similar symptoms

 

[vahegan]

as others called out , we know wifi drop out can originate from either power or router issue. If it was a power issue, reef-pi should love errors indicating system under power, if it was a router issue i would expect other things also to show similar symptoms

Router is rock-solid, nothing else has ever shown any issues.
It was probably the power brick - the capacitors in it might have went dry resulting in higher ripple, that might have went through the second step-down on the Robo-tank board. Don't know. Shouldn't be, but who knows. I have now connected the 60W Meanwell adapter, and re-installed reef pi. Had only one instance since then when the controller was off the WiFi connection and I had to power cycle to bring it back. Will see. If that repeats again, might consider getting a Pi 4 and a wired connection... Pi prices have gone skywards, though, that's why I hesitate.

 

[robsworld78]

Router is rock-solid, nothing else has ever shown any issues.
It was probably the power brick - the capacitors in it might have went dry resulting in higher ripple, that might have went through the second step-down on the Robo-tank board. Don't know. Shouldn't be, but who knows. I have now connected the 60W Meanwell adapter, and re-installed reef pi. Had only one instance since then when the controller was off the WiFi connection and I had to power cycle to bring it back. Will see. If that repeats again, might consider getting a Pi 4 and a wired connection... Pi prices have gone skywards, though, that's why I hesitate.

Maybe it's worth trying a USB to Ethernet adapter. I bought one of those for my Pi Zero W as the wifi was slow, the USB did speed it up a little but not as much as I hoped. Was always able to connect though but I've read the Pi Zero doesn't get along with all routers, maybe it's just something like that.

 

[vahegan]

I considered that option, but I suspect there are other issues with the Pi. I had spent some time trying to figure out why channels #1 & 2 on the power bar did not work. Trying to debug, I disassembled the power bar and traced the signal from the relay through the transistor, and up to the DB9 connector, to find out that it was not present on the connector (when I connected the corresponding pins to another ping with signal, the relay clicked just fine). Since the signal on the connector is around 3V, I suspect that it comes directly from the Pi pins. I re-inserted the Pi on the 40-pin connector on the controller board, suspecting that there might have been a poor contact somewhere, but that didn't help. Therefore I suspect that something is wrong on the Pi and if I am going to replace it, it will probably be a good idea to get a 4: in addition to the wired LAN it also has more resources and I can try connecting a touchscreen and install Kivy. I find that having a dedicated control interface in a fixed position near the tank is more convenient than having to grab the phone or going to the controller - for example, when I want to turn OFF the ATO when I turn down the flow to feed the corals, or want to do a water change. Will see. I'll be traveling by the end of next week, and don't want to do any major changes before I come back in the second week of August.

 

[robsworld78]

I considered that option, but I suspect there are other issues with the Pi. I had spent some time trying to figure out why channels #1 & 2 on the power bar did not work. Trying to debug, I disassembled the power bar and traced the signal from the relay through the transistor, and up to the DB9 connector, to find out that it was not present on the connector (when I connected the corresponding pins to another ping with signal, the relay clicked just fine). Since the signal on the connector is around 3V, I suspect that it comes directly from the Pi pins. I re-inserted the Pi on the 40-pin connector on the controller board, suspecting that there might have been a poor contact somewhere, but that didn't help. Therefore I suspect that something is wrong on the Pi and if I am going to replace it, it will probably be a good idea to get a 4: in addition to the wired LAN it also has more resources and I can try connecting a touchscreen and install Kivy. I find that having a dedicated control interface in a fixed position near the tank is more convenient than having to grab the phone or going to the controller - for example, when I want to turn OFF the ATO when I turn down the flow to feed the corals, or want to do a water change. Will see. I'll be traveling by the end of next week, and don't want to do any major changes before I come back in the second week of August.

Oh no, so sorry for having you go in circles. I think the problem is some missing jumpers explained here. I haven't heard of a Pi GPIO going bad, they seem pretty solid.

Robo-Tank Controller Wiring Diagrams.docx

docs.google.com

You are correct that the 3.3v signal is coming directly from the Pi GPIO's, I did it like that so they can be repurposed. Those hardware backup ports cut the signal for power bar 1, outlet 1 and 2, you can connect any kind of switch to it. These make sure the ports get turned off in case of a Pi or SD card failure if you use a float switch so good for ATO etc. Note all the DC ports have a hardware backup port as well, if no jumper or switch is installed they won't work.

I hear you on the display, have you tried one on the Pi? It is definitely on the slow side but once the browser is open and page loaded it kind of works ok.

 

[vahegan]

Oh no, so sorry for having you go in circles. I think the problem is some missing jumpers explained here. I haven't heard of a Pi GPIO going bad, they seem pretty solid.

Robo-Tank Controller Wiring Diagrams.docx

docs.google.com

You are correct that the 3.3v signal is coming directly from the Pi GPIO's, I did it like that so they can be repurposed. Those hardware backup ports cut the signal for power bar 1, outlet 1 and 2, you can connect any kind of switch to it. These make sure the ports get turned off in case of a Pi or SD card failure if you use a float switch so good for ATO etc. Note all the DC ports have a hardware backup port as well, if no jumper or switch is installed they won't work.

I hear you on the display, have you tried one on the Pi? It is definitely on the slow side but once the browser is open and page loaded it kind of works ok.

Oh, I thought the jumpers were only required on the Lighting & DC module, not? I tried that in an attempt to control a fan directly, used a jumper on DC1 float, but have had issues there, so unconnected the module to try to debug/fix it later. The LEDs on the LAN connectors do not light up when I connect it to the main controller board, right? Because, they do not.
So for the power bar outlets #1 & 2, I also need jumpers on the Backup AC1 & AC2? I didn't realize that, will try it out.

As for the display, I have one, connected to an old RPi2, but it is only a terminal, no GUI on it. I understand that RPi resources may make the GUI slow. Although, I thought that RPi4 has lots of resources and driving a GUI should not be a problem. I mean even the ancient RPi1 (the initial version) had sufficient power for video stream decoding, why should the latest generation be slow in handling a simple graphical interface? Or is it an inefficiency of libraries that are used to draw the interface?

 

[robsworld78]

Oh, I thought the jumpers were only required on the Lighting & DC module, not? I tried that in an attempt to control a fan directly, used a jumper on DC1 float, but have had issues there, so unconnected the module to try to debug/fix it later. The LEDs on the LAN connectors do not light up when I connect it to the main controller board, right? Because, they do not.
So for the power bar outlets #1 & 2, I also need jumpers on the Backup AC1 & AC2? I didn't realize that, will try it out.

As for the display, I have one, connected to an old RPi2, but it is only a terminal, no GUI on it. I understand that RPi resources may make the GUI slow. Although, I thought that RPi4 has lots of resources and driving a GUI should not be a problem. I mean even the ancient RPi1 (the initial version) had sufficient power for video stream decoding, why should the latest generation be slow in handling a simple graphical interface? Or is it an inefficiency of libraries that are used to draw the interface?

The two AC outlets also need them, I added those as they are very useful and not everyone uses DC pumps.

Those lights on the RJ45's should light up, do you have the DC module powered up with 12v or 24v? There's also a power LED on the board, do you see that light up?

The Pi3 or Pi4 has no problem with a display, sorry I was referring to the Pi Zero 2, that one is sluggish.