Mirror turning black after a while

[schlachtkreuzer6]

@retroflex probably we need some code to clear the ram periodically. I think my problem module is the MMM-Globe. But I will have to test it. (about 10 modules… So one module per day… Uff that’s a looooong way.

[cruunnerr]

make two or three per day ;)

than it goes faster

[retroflex]

Maybe this can be the cause?

https://github.com/electron/electron/issues/11739

[schlachtkreuzer6]

@cruunnerr
I’m working 🙄 didn’t test anything yet 😅

[FredJ]

@retroflex I also got the same issue. MMM-resrobot or something it depends on looks to be causing this. Has anyone found a workaround? Resrobot is one of the main reasons I have the mirror so not really happy about disableing it.

[retroflex]

@Alvinger Do you know what could be the problem with ResRobot here?

[Alvinger]

@retroflex I can’t think of anythong off hand.

I am running an older version of MM myself so haven’t researched the electron issues.
Can anyone confirm that the issue started with the new MM version?

As I am running my magic mirror on an original Pi 1 with 256MB RAM with no issues whatsoever I am a bit surprised that the module should cause this.

Edit: MMM-ResRobot has but one dependency and that is the unirest module.

Which version of ResRobot are you running? I did a rewrite to reduce the number of updates w while ago.

[cyberdie]

@alvinger In my case, the problem started after last MM update.

[retroflex]

@cyberdie said in Mirror turning black after a while:

@alvinger In my case, the problem started after last MM update.

Same here. Running the latest ResRobot.

But I think others have reported the same problem but with a different module.

[Alvinger]

@binog @FredJ @cyberdie @schlachtkreuzer6

Looking through the unirest dependencies it looks like it relies on express which may or may not make it sus eptible to the issues with electron.

To resolve this I replaced unirest with node-fetch and pushed the new version of MMM-ResRobot to github.

To upgrade cd to your MagicMirror directory and run the following commands:

git pull
npm remove unirest
npm install

Please post results in this thread!
Hope this resolves the issue.

[retroflex]

@Alvinger Thanks for the quick fix! :)

My first test…

1. “top” without ResRobot shows about 1% CPU when idle.
2. Added ResRobot (old version): “top” shows 100% CPU all the time.
3. Updated ResRobot, removed unirest, npm install: “top” now shows 1% CPU when idle. Yay!

Looks good so far. Too early to tell anything about memory leaks, but I will keep my mirror running and check tonight. Currently electron shows 18% mem used (for the 2 electron processes).

[retroflex]

Short lived happiness it seemed… After a few minutes CPU is now back at 100% and memory use increasing. :(

Also tested after reboot of my Pi3 with the same results.

This is my config:

		{
			module: "MMM-ResRobot",
			position: "top_right",
			header: "",
			config: {
				routes: [
					{from: "740015970", to: "740015568"},
				],
				updateInterval: 2 * 60 * 1000,	// Update every second minute.
				maximumEntries: 6,  // Number of departures to show on screen
				truncateAfter: 5,   // A value > 0 will truncate direction name at first space after  characters. 0 = no truncation
				apiKey: "*******"  // Your ResRobot apiKey
			}
		},

[Binog]

But still, I am not using ResRobot,. and do have the same problem, from time to time,…

[Alvinger]

@retroflex
I will dig up another respberry pi and try it out. My test environment is a dietpi vm which runs on a pc so I wouldn’t expect the cpu load to be representative of the load on the pi.

[FredJ]

I also dont think the issue is with ResRobot itself. But it is great that you are looking into it @Alvinger

[cyberdie]

no ResRobot running on my setup. Think the problem is electron itself.

[Alvinger]

I agree with others that the problem is most likely related to Electron and not module-specific but rather related to which module dependencies it has.

[thejoaovitor]

A solution, (not perfect) is to use MMM-WatchDog, it keeps an eye on your UI and quits app in case the UI crashes, combined with the PM2 process manager, MM2 will automaticly restart after a UI failure.

https://github.com/MichMich/MMM-WatchDog

[Kuhlemann]

The problem still seems to persist also with the newest version of MM. Maybe someone has found a real solution?
I will try out MMM-WatchDog meanwhile.

[dazza120]

@kuhlemann a cron job is the only way think it seems that it taxes the Pi too much after 8hours or so hence Cron Job every 8hours or so