Node-Red on RPi restarts after ~2 weeks

General
21

I would start by removing the following nodes (in order, not at the same time)

  • calculate
  • smooth
  • counter

The first 2 store data in an internal array for later computation

In fact, looking at the src of the calculate node, it (for some reason) stores the entire msg object (even though it only appears to use payload in computing the averages!).

1 Like
22

Also, do you actually use play audio on all of them?

23

@Steve-Mcl i just deactived these three one after one. Ram was at 34%, after deactivating it rises up to 47% (in 3 minutes)

@Colin I don´t use play audio.

Could it theoretically be a hardware defect of the ram?

24

I use only one calculate node in this flow.
in my other node-red I use > 10 without problems.

grafik
grafik234×184 3.55 KB

the calculate gets every 2s this msg.object

in my other node-red the calculate gets much more crap in msg.object

I use only one smooth node in this flow, in my other node-red I use > 20 without problems.
it´s 100% the same input like in my other node-red:

grafik
grafik305×882 10.6 KB

every hour max. 35 msg

and I also use only one counter node, which has actually not been used since the reboot.

25

The only major difference to the other systems and my last idea is the following:

boot/config.txt add:

[all]
gpio=4,5,7,11,17,22,23,24,25,26,27=op,dl
gpio=6,8,9,10,12,13,16,21=ip,pu

and some gpio nodes:

grafik
grafik542×356 9.69 KB

with initialization

Can this be ruled out as a reason?

next step: 2nd RPi as a pure simulator (without modbus)

is it completely impossible to see a detailed RAM listing in the Node Red core?

26

It isn't going to be a fundamental config issue, it is what is being done with the data.

You showed an installed node list earlier, but said that some of them are unused. I would start by removing all those that are not used and making sure all the rest are up to date.

You showed earlier you were getting regular modbus timeouts. Do the other systems suffer from this too? It is certainly not impossible that something like this could find a memory leak that has not been noticed previously.

Next time you restart could you show us the full node red startup log please.