swappiness to 0 doesn't seems to work

CharlesErickT · Dec 1, 2017

Hello,

I've recently changed the swappiness on my host to 0 since there is plenty of ram available. Rebooted to apply the change.

Now my container still consume swap even when there is memory free.

Here's in the webui:

Here's the free output inside the container:

Code:

              total        used        free      shared  buff/cache   available
Mem:           1024          33         583        4363         406         583
Swap:           256         190          65

Maybe I'm reading the output wrong but it seems to have enough ram to not swap ?

Here's my pveversion (Running 5.1 with ZFS)

Code:

pve-manager: 5.1-35 (running version: 5.1-35/722cc488)
pve-kernel-4.13.4-1-pve: 4.13.4-25
libpve-http-server-perl: 2.0-6
lvm2: 2.02.168-pve6
corosync: 2.4.2-pve3
libqb0: 1.0.1-1
pve-cluster: 5.0-15
qemu-server: 5.0-17
pve-firmware: 2.0-3
libpve-common-perl: 5.0-20
libpve-guest-common-perl: 2.0-13
libpve-access-control: 5.0-7
libpve-storage-perl: 5.0-16
pve-libspice-server1: 0.12.8-3
vncterm: 1.5-2
pve-docs: 5.1-12
pve-qemu-kvm: 2.9.1-2
pve-container: 2.0-17
pve-firewall: 3.0-3
pve-ha-manager: 2.0-3
ksm-control-daemon: 1.2-2
glusterfs-client: 3.8.8-1
lxc-pve: 2.1.0-2
lxcfs: 2.0.7-pve4
criu: 2.11.1-1~bpo90
novnc-pve: 0.6-4
smartmontools: 6.5+svn4324-1
zfsutils-linux: 0.7.2-pve1~bpo90

Right now we need to 'swapoff -a' and 'swapon -a ' every couples of day to fix the issue temporary

Any idea idea what could cause this ?
Thanks

dietmar · Dec 1, 2017

You need to set swap to 0 on the container Resources/Memory configuration.

CharlesErickT · Dec 1, 2017

dietmar said:
You need to set swap to 0 on the container Resources/Memory configuration.

That make sense. Would the best practice when using ZFS be to keep swappiness on the host to 10 and disable it for the container or it doesn't matter and I could set it at 0 on the host while disabling it for the container ?

Rhinox · Dec 1, 2017

IMHO the best practice is to disable swap for both PVE-hosts and VMs/containers completely. For a few years one of the things I do on every PVE-host is "zfs destroy rpool/swap" (not just "swappiness 0"). Never had any problem. And when I'm talking about it, swap on ZFS is still not very stable solution. One more reason to get rid of it completely...

LnxBil · Dec 2, 2017

Please also consider to use zram, which creates a compressed swap image in your ram - yes sounds strange, but works perfectly!

CharlesErickT · Dec 4, 2017

Rhinox said:
IMHO the best practice is to disable swap for both PVE-hosts and VMs/containers completely. For a few years one of the things I do on every PVE-host is "zfs destroy rpool/swap" (not just "swappiness 0"). Never had any problem. And when I'm talking about it, swap on ZFS is still not very stable solution. One more reason to get rid of it completely...

Yeah this is what I'll do because we have plenty of ram available and so far the swap only brought issues so disabling it seems like the logical solution. Thanks for your input

LnxBil · Dec 4, 2017

CharlesErickT said:
Yeah this is what I'll do because we have plenty of ram available and so far the swap only brought issues so disabling it seems like the logical solution.

Filled up swap is not the root cause of your problem - swap is only used if there is not enough ram available, so if you disable swap and you run out of ram, processes will be killed, so this is absolutely the wrong thing to do. Increase your ram to the machines and keep the swap setting. Monitor what your containers/vms do and react accordingly.

CharlesErickT · Dec 4, 2017

LnxBil said:
Filled up swap is not the root cause of your problem - swap is only used if there is not enough ram available, so if you disable swap and you run out of ram, processes will be killed, so this is absolutely the wrong thing to do. Increase your ram to the machines and keep the swap setting. Monitor what your containers/vms do and react accordingly.

Yes I'm aware of that but based on what I've posted originally it's like the container is swapping even if the swappinness is 0 and there is plenty of ram available. We have some monitoring software in place to monitor the ram so it should never be full and if it even gets high we can act accordingly.

LnxBil · Dec 4, 2017

CharlesErickT said:
Yes I'm aware of that but based on what I've posted originally it's like the container is swapping even if the swappinness is 0 and there is plenty of ram available. We have some monitoring software in place to monitor the ram so it should never be full and if it even gets high we can act accordingly.

Yes, should work and if something gets oom-killed, memory can be increased further. Sometimes, simple processes like locales generation can need a lot of temporary memory, so those bottlenecks really occur often for a very, very short amount of time. Same for downloading and parsing package update lists etc.

I only want to cover all bases so only a small followup question:
Are you sure that you monitor the container ram correctly? Depending on the guest operating system and monitoring software, there have been problems in the past with different versions, that were also discussed in forums.

CharlesErickT · Dec 4, 2017

LnxBil said:
Are you sure that you monitor the container ram correctly?

I'm using Nagios with a custom script that reads /proc/meminfo of the container. The Nagios report is consistent with what I'm seeing in the Proxmox webui.

Here's free -m from a container

Code:

              total        used        free      shared  buff/cache   available
Mem:           1024          39         922        4820          61         922
Swap:             0           0           0

And Nagios report:

Code:

OK - Total: 1024 MB - Free: 988 MB - 4% used : Swap: 0 MB - Free: 0 MB - 0% used

Which seems to work fine

LnxBil · Dec 4, 2017

Yes, that's fine. Let it run it a couple of weeks and please report back if it solved your problem(s).

CharlesErickT · Dec 4, 2017

LnxBil said:
Let it run it a couple of weeks and please report back if it solved your problem(s).

Will do ! Thanks for your help

re-host.eu · Dec 5, 2017

CharlesErickT said:
Yes I'm aware of that but based on what I've posted originally it's like the container is swapping even if the swappinness is 0 and there is plenty of ram available. We have some monitoring software in place to monitor the ram so it should never be full and if it even gets high we can act accordingly.

I had the same issue... For the swappiness setting you need to reboot the host system.
So im using Swappiness 0

CharlesErickT · Dec 5, 2017

re-host.eu said:
I had the same issue... For the swappiness setting you need to reboot the host system.
So im using Swappiness 0

As mentionned in my OP, I did reboot to apply the changes but to no avail.

re-host.eu · Dec 5, 2017

CharlesErickT said:
As mentionned in my OP, I did reboot to apply the changes but to no avail.

What is the output of

Code:

cat /etc/sysctl.conf | grep vm.swappiness

?

CharlesErickT · Dec 5, 2017

Here's the ouput:

Code:

vm.swappiness = 0

re-host.eu · Dec 5, 2017

ok, i thought maybe the swappiness setting might got lost.. but this isnt the situation. Strangely for me it works ..

CharlesErickT · Dec 5, 2017

Yeah this is weird, maybe a bug in the version of Proxmox I use. Thanks for your help anyway

Search

Search

swappiness to 0 doesn't seems to work

CharlesErickT

Member

dietmar

Proxmox Staff Member

CharlesErickT

Member

Rhinox

Active Member

LnxBil

Distinguished Member

CharlesErickT

Member

LnxBil

Distinguished Member

CharlesErickT

Member

LnxBil

Distinguished Member

CharlesErickT

Member

LnxBil

Distinguished Member

CharlesErickT

Member

re-host.eu

New Member

CharlesErickT

Member

re-host.eu

New Member

CharlesErickT

Member

re-host.eu

New Member

CharlesErickT

Member