Proxmox using swap, the everlasting thread ...

[skraw]

Hello all,

We are using kernels 5.0.15-1-pve and set swappiness globally to 0 by sysctl and in all processes by using this script

#! /bin/bash
#
# set swappiness to
SW=0

/usr/sbin/sysctl vm.swappiness=${SW}
# fix for proxmox kernel swappiness problem, yes, it must be executed twice
for i in `/usr/bin/find /sys /var/lib 2>/dev/null | /bin/grep slice | /bin/grep swappiness` ; do /bin/echo ${SW} >"${i}" ; done
for i in `/usr/bin/find /sys /var/lib 2>/dev/null | /bin/grep slice | /bin/grep swappiness` ; do /bin/echo ${SW} >"${i}" ; done

exit 0

(Taken from another thread around this topic)

On a host with few kvms this works as expected, (h)top shows zero used swap.
But on a host with around 20 kvms things are different. (h)top shows a small amount of swap used (around 9,3 M), but /proc shows this:

# grep VmSwap /proc/*/status | grep -v " 0 kB" | sort -k 2 -n -r
/proc/6552/status:VmSwap: 178924 kB
/proc/14203/status:VmSwap: 48212 kB
/proc/29332/status:VmSwap: 38884 kB
/proc/23862/status:VmSwap: 30824 kB
/proc/2143/status:VmSwap: 17904 kB
/proc/42786/status:VmSwap: 15084 kB
/proc/29740/status:VmSwap: 12092 kB
/proc/31301/status:VmSwap: 7840 kB
/proc/18264/status:VmSwap: 7692 kB
/proc/34265/status:VmSwap: 6440 kB
/proc/9954/status:VmSwap: 6392 kB
/proc/35872/status:VmSwap: 5908 kB
/proc/27786/status:VmSwap: 5744 kB
/proc/7013/status:VmSwap: 5548 kB
/proc/39290/status:VmSwap: 5340 kB
/proc/36510/status:VmSwap: 2272 kB
/proc/1772/status:VmSwap: 1932 kB
/proc/17836/status:VmSwap: 1436 kB
/proc/12571/status:VmSwap: 1396 kB
/proc/38452/status:VmSwap: 1144 kB
/proc/35166/status:VmSwap: 1012 kB
/proc/5010/status:VmSwap: 848 kB
/proc/15672/status:VmSwap: 652 kB
/proc/25142/status:VmSwap: 616 kB
/proc/4832/status:VmSwap: 496 kB
/proc/9891/status:VmSwap: 440 kB
/proc/3561/status:VmSwap: 404 kB
/proc/12441/status:VmSwap: 400 kB
/proc/2142/status:VmSwap: 364 kB
/proc/26728/status:VmSwap: 328 kB
/proc/22863/status:VmSwap: 272 kB
/proc/2144/status:VmSwap: 148 kB
/proc/15585/status:VmSwap: 80 kB

These are almost all kvms. According to my understanding this adds up to a lot more than 9M.
Here is top:

top - 14:18:42 up 148 days, 20:20, 6 users, load average: 3.41, 2.93, 3.08
Tasks: 678 total, 1 running, 677 sleeping, 0 stopped, 0 zombie
%Cpu(s): 6.5 us, 4.2 sy, 0.0 ni, 88.8 id, 0.0 wa, 0.0 hi, 0.4 si, 0.0 st
MiB Mem : 128880.3 total, 58718.4 free, 68338.0 used, 1823.8 buff/cache
MiB Swap: 7168.0 total, 7158.7 free, 9.3 used. 58613.9 avail Mem

As you can see there is no reason for swapping as almost half the mem is free.
If I cannot tune this with swappiness, what else can we do?
--
Regards
Stephan

 

[wolfgang]

Hi,

swappiness=0 does not mean you disable swap.
If you like to disable swap you have to remove the swap partition

swappiness
==========

This control is used to define how aggressive the kernel will swap
memory pages.  Higher values will increase aggressiveness, lower values
decrease the amount of swap.  A value of 0 instructs the kernel not to
initiate swap until the amount of free and file-backed pages is less
than the high water mark in a zone.

The default value is 60.