[Detector Support]: M.2 A+E Key Single TPU and 5k Major Page Faults per Second. #15777
Unanswered
reefland
asked this question in
Detector Support
Replies: 1 comment 3 replies
-
It's unclear that this is actually related to something Frigate can address. A a "major fault" refers to a page fault that requires the OS to retrieve data from the disk to satisfy a memory access request. There could be many reasons for this, but all that's being indicated is that there's a lot of transfer between RAM and swap space. |
Beta Was this translation helpful? Give feedback.
3 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Describe the problem you are having
I did not have this issue using the USB detector. There is no obvious issue, Frigate is working fine. I would not know there was an issue if Prometheus hadn't raised the alert. The Prometheus metric is
node_vmstat_pgmajfault
, which identifies the computer with the issue. I then usedTop
, adding fieldnMaj
and sorting by it. The value increases by 5k a second on each of thefrigate.capture
processes.Prometheus alert:
Memory major pages are occurring at very high rate at 192.168.10.218:9100, 500 major page faults per second for the last 15 minutes, is currently at 5071.13.
The alerts triggers at 500 per second, I'm over 5000.I have given Frigate
memory: 4608Mi
resource limit and it is using less than 1/2 of that:Version
0.14.1
Frigate config file
docker-compose file or Docker CLI command
Running under K3s.
Relevant Frigate log output
Install method
HassOS Addon
Object Detector
Coral
Screenshots of the Frigate UI's System metrics pages
Any other information that may be helpful
Installed using Helm Chart. Have been using Frigate for years with same config with USB TPU without issues. I removed the USB and switched to M.2 A+E single TPU detector and updated the
detector
section of the config.Beta Was this translation helpful? Give feedback.
All reactions