ZoneMinder Forums

iconnor wrote: ↑Fri May 14, 2021 12:24 pm You havn't been reading along.
Set your imagebuffercount to 3.

Sorry, it can be easily, that I miss something. There are a lot of information regarding ZM and at the very start it's not so easy to find out, what is relevant to your case and to understand this information. Now I see, that it was already mentioned in the topic "Features in 1.35.*" and in the Option Help.

Thank you for the tipp! I've set the imagebuffercount to 3 and now the memory consumption has become much lesser.

I tried to remove the limits by setting Maximum Image Buffer Size to 0. Memory usage quickly went wild, so here's some log files (at level 4, let me know if you'd like me to re-run at level 3).

There is no change to the lacing. (De-interlacing is off)

https://drive.google.com/file/d/144lbRe ... sp=sharing

https://drive.google.com/file/d/1ls29JQ ... sp=sharing

https://drive.google.com/file/d/1PDODKe ... sp=sharing

So for a 12fps keyframe interval, I would recommend a max image buffer setting of like 25 to 30.

The first two logs show reasonable behaviour.
The third shows several things:
1 it is using cuda to decode. You might try without. hwaccel has really not worked out in practice
2 decoding is not keeping up. Your packetqueue is growing forever. Set a max image buffer.

I think if you look at top (Hit M to sort by ram use, c to show command line) you will find that zmc -m 2 is consuming all the ram and not the others.

It is taking FOREVER to convert that large image from the nv12 format that cuda uses to rgb0 that we use.

Turn off cuda. I think we will get better results just doing regular software decoding.