PVE node SWAP usage - vm.swappiness configuration script

[heinemannj]

🌟 Briefly describe the feature

Add support for vm.swappiness configuration

📝 Detailed description

To apply vm.swappiness change on a PVE node immediately, but not persistant:

sysctl -w vm.swappiness=0

For persistant change modification of /etc/sysctl.conf:

cat <<EOF > /etc/sysctl.conf
vm.swappiness = 0
EOF

💡 Why is this useful?

PROXMOX VE node default settings for SWAP usage:

Size: 8GB
vm.swappiness=64

Results in regular Netdata alerts because the PVE node SWAP usage is above 98%.
I think almost every Promox HomeLab user is facing this kind of issues too.

After the change the swap usage in most of my LXCs is under 0.1%.
https://forum.proxmox.com/threads/swap-usage-100.109491/post-780661

Only a few bestie LXCs are consuming much more (1-100%).

[MickLesk]

Thanks for the suggestion. However, this is better suited as a documentation/guide rather than a script feature.

main Reasons:

• This is a PVE host-level configuration, not container-related
• swap configuration is highly subjective and depends on individual use cases
• Its a simple 2-command change that doesn't warrant automation
• Some users prefer no swap at all

me as example, ive removed swap completly from my main node, because it caused more problems than it solved. Only my DEV node still runs swap to test scripts that require it.

[heinemannj]

Based on my investigations:

• PVE host-level configuration are container-related as well - clearly seen in most of my LXCs ...
• Having an automated question and answer game will simplify the configuration for not so experienced users

Second point:

• The mayority of our users are searching in Proxmox VE Helper-Scripts Web-GUI first and most likely NOT in the discussion areas ...

Question:
How to removed swap completly from a PVE node?

[MickLesk]

with swapoff -a

My concerns are here:

1. vm.swappiness is highly environment-specific - what works for one setup may cause issues for another. The "right" value depends on workload, available RAM, storage type (SSD vs HDD), and use case.

2. Host-level configuration with significant impact - automating this could lead inexperienced users to change critical system settings without fully understanding the implications.

I think this would be better served as documentation or a discussion post with:

• Explanation of what vm.swappiness does
• Different scenarios and recommended values
• How to monitor swap usage properly
• When to add more RAM vs. tune swap settings

That said, I'd be interested to hear what other @community-scripts/contributor think about this. If there's broader consensus that this would be valuable as an automated feature, I'm open to reconsidering.

[heinemannj]

swapoff -a on an PVE node is persistant?

"... This is a PVE host-level configuration, not container-related ..."

• Yes - you are right:

root@esphome:~# sysctl vm.swappiness
vm.swappiness = 60

"... That said, I'd be interested to hear what other @community-scripts/contributor think about this. If there's broader consensus that this would be valuable as an automated feature, I'm open to reconsidering. ..."

OK - make sense - How to proceed?
Open a new discussion or let this case open to discuss here?

[tremor021]

swapoff -a disables it right away, but its not persistant. You will need to edit fstab for it to stay that way. But as most of PVE nodes stay running until they die, i can see why swapoff -a can be a solution to some.