LXC Guest not clearing memory buffers/cache

[sg86]

Hi, so I'm running "zoneminder" software on a LXC container, I've assigned it plenty of RAM which i keep increasing because its swapping.

Digging further i see that it is incorrectly reporting actual RAM usage, it is sitting (in HTOP), at 27GB / 22 GB


"free -hm" shows only 10mb free, 22GB of the usage is sitting in buffers/cache.

              total        used        free      shared  buff/cache   available
Mem:            22G        1.1G         10M        6.6G         20G         10M
Swap:          4.0G          0B        4.0G

When i stop zoneminder it sits like below forever seemingly.

              total        used        free      shared  buff/cache   available
Mem:            22G        719M        6.4G        821M         14G        6.4G
Swap:          4.0G          0B        4.0G

Now I have spoken with the zoneminder developers and they say this is an issue with LXC not being able to clear its buffers, is there something i can do? Am i able to make proxmox clear these when they reach dangerous levels?

What is also quite annoying is proxmox "summary" UI shows that only 4-5GB of RAM are used, not the 22GB shown on this single guest

 

[keith123]

I am having the exact same issue as you have mentioned. also looking for a fix.
Have you found anything yet yourself?

 

[keith123]

Hello,

I have made a temporary work around for myself, if its any use for you. it works fine for my ZoneMinder.

when I run
"cat /proc/meminfo" I see all the used memory in Cached never clears.
I run "sync; echo 1 > /proc/sys/vm/drop_caches" to clear the PageCache and see my ZoneMinder vM in proxmox clear up 3 GB of RAM.

I now just have a cron task to run every 2 hours to execute "/bin/sync; echo 1 > /proc/sys/vm/drop_caches" and no longer have the issue.
Still its just a hacky work around

 

[JasonBorn]

did you solved that problem? I got the same issue

 

[mathx]

did you solved that problem? I got the same issue

Wonder if its fixed in 6.x

 

[mathx]

It's not fixed in 7.1 and you cant drop caches in a CT, only for the whole host. On my 256GB host with 96GB of arc, that seems like a massive cache loss/performance hit.

#echo 1 > /proc/sys/vm/drop_caches
bash: /proc/sys/vm/drop_caches: Read-only file system

These are basically the same thread/issue, please merge:

https://forum.proxmox.com/threads/c...-usage-until-oom-killer-kill-processes.67666/

 

[mathx]

You may not want or need to drop caches -- 'buffers/cache' is also charged for TMPFS usage (misleadingly by LXCFS) - and tmpfs's cannot be evicted from ram.

What's using all the tmpfs ram? In /run we have systemd-journald logs. And systemd provides no sane default for vacuuming logs.

see https://forum.proxmox.com/threads/c...-oom-killer-kill-processes.67666/#post-472649 for details.

 

[mm553]

Same issue here:
https://forum.proxmox.com/threads/lxc-and-tvheadend-continuity-counter-error.97726/#post-486743