Is it typical to see the resident memory of allmon3 creep up and up over time? Another system admin on our end is questioning the RAM usage of allmon3.
I have two rpi3 running raspbian trixie + ASL3. Both have uptimes of ~ 2 days and the RES ram for allmon3 is at a crazy 412MB and 452MB and seems to keep climbing.
Both pi’s have light to no allmon3 actual usage (the web page is not particularly used).
pi3 (1)
********** AllStarLink [ASL] Version Info **********
OS : Debian GNU/Linux 13 (trixie)
OS Kernel : 6.12.62+rpt-rpi-v8
Asterisk : 22.7.0+asl3-3.7.1-1.deb13
ASL [app_rpt] : 3.7.1
AFAICS the memory usage will step up for each logged in monitoring console. The amount of resident memory will fluctuate with the system state, normally.
Valgrind shows no memory leaks in the current version of allmon3.
Hmm. It would be safe to say that there are zero or one logged in users max on these two devices at any time for allmon3. If there are any users of Raspbian Trixie + ASL3 + allmon3 that can say what their RES values for memory are after a few days that would be interesting.
I see anywhere from 500k to 800k bytes of virtual memory used after a week or two, so I don't worry as long as the resident memory is <= that value.
Just because the number looks big doesn't mean it's necessarily a worry. You may have noticed that allmon3 uses Python, so what's loaded in memory in part is whatever the language pulls in during execution.
Let us know if you observe an unbounded memory usage increase.
I have seen unreasonable memory inflation if you have someone on the internet who does a drive-by on the websockets and blast the crap out of them. Something that was doing that to Node 2004 a week or so ago. Finally went away.
Allmon3 is all Python which is not a memory-lightweight. IT also stores the entire “Allmon DB” in memory.
That looks like you just fired up and have one login if I'm reading the tea leaves correctly. I see 800k/400k, more like your first posting, on an RPi5 after 24 hr and ~1000 keyups on a net with 100-250 nodes.
Fortunately I have no pathologically connecting nodes, so @N8EI's caveat does not apply.