[Support]: System unresponsive due to memory filled up completely #9936
Replies: 3 comments 6 replies
-
I would suggest using preset-vaapi instead of manual hwaccel args can see a number of camera timeouts as well |
Beta Was this translation helpful? Give feedback.
-
I've configured now a Docker resource limit and reservation on memory and cpu |
Beta Was this translation helpful? Give feedback.
-
Allright, this seems like my issue as well - but I have different errors than here, but symptom with setup is same #10044 |
Beta Was this translation helpful? Give feedback.
-
Describe the problem you are having
Issue: Unresponsive system due to memory being filled completely.
In the memory picture above you'll see the capture of an already active crash on the left part.
At 9:39 I increase the memory from 4GB to 6GB in Proxmox
followed by a memory decrease to normal level and small system recovery window.
At 9:59 and a 2nd crash, filling up the memory again, and I increase the memory again, with repeating results
Version
0.13.1-34FB1C2
Frigate config file
Relevant log output
FFprobe output from your camera
Frigate stats
Operating system
Proxmox
Install method
Docker Compose
Coral version
USB
Network connection
Wired
Camera make and model
Voortuin: SJCAM, Deurlamp: EZVIZ, Deurbel: Reolink
Any other information that may be helpful
I'm receiving body is too long (maximum is 65536 characters) from github, slimming down my feedback and editing after.
It doesn't matter how much memory head-room is given to the system, it gets filled to the max in no time.
After I notice Frigate is not working anymore (no web service, not even console access through Proxmox),
I add a bit more memory, and the system is responding again.
The memory usage goes back to a normal level, and the system continues to work and I can decrease the memory.
Sometimes the system works perfectly normal at about 1GB of RAM. But when the issue occurs, it doesn't matter how much RAM you have configured, it's been occupied in no time, with an unresponsive system as result.
I've installed Frigate on Proxmox LXC with Turnkey Linux and Docker, having the following resources:
Hardware: NUC: D34010WYK - 4 x Intel(R) Core(TM) i3-4010U CPU @ 1.70GHz with 16GB RAM.
Proxmox LXC container:
The GPU and USB Coral are passed-through Proxmox and Docker.
Coral has ~ 10ms inference speed, and very low load.
The Intel-vaapi GPU has as well a very low usage
The NUC is running Home Assistant, DNS, and a few lightweight processes in separate LXC's or VM's.
The average CPU load is ~ 50%, with only spikes as Frigate halts.
Beta Was this translation helpful? Give feedback.
All reactions