Memory leak?

[steviehs]

From what I see, there are some open issues in the original project, which should now be resolved here. One is a memory leak (hey, wasn't that the reason for golang?): If I have octoscreen running and the printer is disconnected, I see the screen "waiting for connection" - which has do be discussed in another issue - still after some hours, even octoprint get's killed as octoscreen is eating up more and more memory on my RPI3.

[alexkv]

Does it happen only when OctoPrint is disconnected? How fast does octoscreen eat memory on your environment?

[alexkv]

Could you check this behaviour with a new version https://github.com/Z-Bolt/OctoScreen/releases/tag/v2.2.0?

[steviehs]

I don't want to run the printer over night just connected. So it happened without the printer connected in about 8 hrs over night on a rpi3b. I will try the new version when I am back home.

[steviehs]

Ok, checked it now - and it is the same behaviour. I guess you can easily reproduce it: just let octoscreen and octoprint run for 1-2 days without having the printer on/connected. After that ocotoscreen has eaten up so much memory, that a rpi3b+ won't have enough mem for octoprint anymore...

 PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                       
18549 root      20   0 1056732  56672  43064 S   2.0  6.3  29:07.21 OctoScreen        

[MatthewCroughan]

@steviehs this has to be an issue with Octoprint itself, no? It's the serial requests to it that make Tornado run out of memory. Octoscreen is causing the memory leak in Octoprint.

[MatthewCroughan]

also, do you think you could link those issues in Octoprint you were referring to?

[steviehs]

Checking over night, Octoprint does not use more memory, but OctoScreen does. So I would see this as an issue of OctoScreen.
Octoprint is running for months with the printer shut down, but OctoScreen dies approx after 15 hr or so on a rpi3b+

[MatthewCroughan]

@steviehs For me, as I've linked in the issue, it's clearly Octoprint that's causing a memory leak. Have you got logs? Take a look at mine in #58

[steviehs]

As said, it is not Octoprint growing in mem usage but Octoscreen. So maybe you have another issue. I am from my side not really looking into Octoscreen at this point, as long as I do not see any improvements to this bug. I do not have time nor understanding to trace this bug in go, so unless anybody else is looking at it I have to wait.

[MatthewCroughan]

I think I understand what you're saying. You believe that Octoscreen grows in memory usage, to the point at which Octoprint is unable to allocate memory? The problem is that when I do free -m I don't see a reason for this, both applications should be able to live, there's always 500M free.

[MatthewCroughan]

@steviehs Issue 56 explains why I believe that this is the fault of Octoprint. Because restarting Octoprint fixes the issue. The state of Octoscreen does not matter.

@alexkv Just to add to previous discussion, this happens for me on a Pi 3B within 1 hour. There are around 200 request being made every minute.