[SOLVED] qemu high memory consumption (and stuck)

[FischBrei]

Hi,

I am using a raid1 on my host system to store data. I try to share data from the host to the vm.
At the moment I am trying to share it with disk mount:

I have added the following line to the qemu config file of VM1:

scsi1: /dev/md0

Afterwards I booted the vm and mounted the dev with

mount UUID=XXXXXXXXXXXXX /mnt

Works like a charm. Then I have tried to copy some of the mounted data to the local vm disk

cp -R /mnt/testdir/ .

The memory consumption raised to 96%. After the copy was done, it stayed at 96%.... I had to shutdown the VM to free up the resources.


Another weird behavior:
I have stored some data on the local vm disk and fired up the samba server.
I accessed a share over the network with a Windows client and transfered around 20GB form the vm.

Same behavior: The memory consumption raised to 96%. After the copy was done, it stayed at 96%.... I had to shutdown the VM to free up the resources.


I also tried to share the data from the host via NFS server, but as soon as the vm has some high access/network transfers, the memory consumption will raise and stay till a reboot happens.

When I do this with a container, everything works fine. I can create mount points and access or copy the data without the weird memory behavior.

I have the latest version of pve.

Does someone have a clue whats happening or how this can be fixed?
Can someone confirm this behavior?

Host: HP Gen8 with 12GB RAM, PVE on SSD and 2 HDD as RAID-1
VM1: Ubuntu Server 16.04 with 4GB RAM
VM2: Debian 9 with 4 GB RAM
CT: Ubuntu 16 2GB RAM

 

[FischBrei]

So when I clear the cache by executing the following command, the memory is back zo a good state (~5% used):

echo 1 > /proc/sys/vm/drop_caches

So now I am looking for qemu cache optimization(?) or does someone know the reason of this bahavior?

I am just curious is this behavior normal for qemu or for proxmox? Or is my hardware somewhat damaged or my settings are incorrect?

With the current situation it is not possible for me to use proxmox/qemu. And I regret my decision buying promox for now.

 

[fabian]

you might want to look at https://www.linuxatemyram.com/

 

[FischBrei]

Ok, thank you for the informations.

Even it is not a problem with the cache or linux at all, but it is still weird in the overview and can easily confuse a user. Maybe it is a good idea to adapt the overview accordingly to your link. So it will not display "Please panic your vm is using 98% of your ram"... :- )

 

[Patlechriss]

Hello

If this can help someone;

I had an issue with extreme high memory load on idle. Enabling I/O threads on my VM disks resolve my issue. Don't know if its related.

proxmox-ve: 5.2-2 (running kernel: 4.15.17-2-pve)
4 nics, 5 bridges, 1 bond (lacp)
10 vm (all debian)
2 ssd + 4 disks raid 10 (zfs) on an areca 1210 controller (lun mode)
32 G ram

Pat

 

[Patlechriss]

Please ignore previous post, it doesn't work.

Status

CPU usage

1.33% of 4 CPU(s)
Memory usage

96.75% (15.48 GiB of 16.00 GiB)
Bootdisk size

32.00 GiB

Sorry...

 

[inov]

What is the solution?