Reducing swap usage on a hypervisor host to protect SSD lifespan

The Linux kernel’s default vm.swappiness (60 on most distributions) makes the kernel start swapping out fairly aggressively, well before physical memory is actually exhausted. On a hypervisor host with plenty of RAM, this can mean noticeable and unnecessary swap-to-SSD write traffic — which shortens SSD lifespan over time for no real benefit.

How it works

vm.swappiness is a 0–100 kernel tunable controlling how aggressively the kernel prefers to swap out anonymous memory pages versus reclaiming page cache:

• vm.swappiness=0 — swap only when physical memory is nearly exhausted.
• vm.swappiness=10 — swap once free memory drops to roughly 10%.
• vm.swappiness=60 (typical default) — swaps considerably earlier and more readily.

Steps

1. Check the current value

cat /proc/sys/vm/swappiness

2. Set a new value

sysctl vm.swappiness=0

Use 0 to swap only under real memory pressure, or a low value like 10 for a small safety margin. Make it persistent across reboots by adding it to /etc/sysctl.conf or a file under /etc/sysctl.d/:

vm.swappiness=0

3. Clear out existing swap usage (optional)

Applying the new swappiness value doesn’t undo memory already swapped out. To force everything back into RAM immediately:

swapoff -a   # can take several minutes on hosts with a lot of swapped data
swapon -a

Verify

cat /proc/sys/vm/swappiness

Troubleshooting

• swapoff -a seems to hang: it can legitimately take on the order of 10–15 minutes on a host with a large amount of data currently in swap, since every swapped page has to be paged back into RAM before the swap area can be released — this is expected, not a fault.

Related

• Removing the Proxmox VE subscription notice — another small generic Proxmox VE host tweak from the same source.
• Debian — another small generic Proxmox VE host tweak.

Sources

• Legacy wiki.js: Proxmox helper tweaks (private)