Many Linux administrators face this scenario: your monitoring system keeps sending swap usage alerts even when RAM becomes available. The swap space remains stubbornly allocated, triggering hourly notifications that clutter your inbox while providing no actionable information.

Linux kernel's swappiness behavior keeps swapped-out pages in disk storage until absolutely necessary to move them back to RAM. This conservative approach prevents unnecessary disk I/O but leads to our alert problem.

# Check current swappiness value (default is usually 60)
cat /proc/sys/vm/swappiness

# View current swap usage
free -h
swapon --show

Here are three effective methods to free swap without system restart:

Method 1: The Systematic Approach

This method safely moves pages back to memory when possible:

# Make kernel attempt to swap in pages
sudo swapoff -a && sudo swapon -a

Note: This briefly disables all swap, so ensure you have sufficient free RAM.

Method 2: Targeted Page Clearing

For more granular control, use this advanced technique:

# Find processes using swap
sudo smem -s swap -r | head

# Then clear page caches (as root):
echo 3 > /proc/sys/vm/drop_caches

Method 3: Swappiness Adjustment

Temporarily reduce swappiness to encourage pages back to RAM:

# Set swappiness to 10 temporarily
sudo sysctl vm.swappiness=10

# Let it work for a while, then check swap
sleep 300; free -h

Create a cron job to prevent alerts during low-usage periods:

# In /etc/cron.hourly/swap_cleaner
#!/bin/bash
[ $(free | awk '/Swap:/ {print $3}') -eq 0 ] && exit 0
[ $(free | awk '/Mem:/ {print $7}') -gt 2097152 ] && swapoff -a && swapon -a

Sometimes the better solution is tuning your alerts:

# Example Nagios check that considers free memory
command[check_swap]=/usr/lib/nagios/plugins/check_swap -w 50% -c 20% -K

• Always maintain some swap for emergency situations
• Test these commands in non-production first
• Consider adding buffer to your alert thresholds
• Document any changes to swappiness values

---

Many Linux admins encounter situations where swap remains fully utilized even when system RAM becomes available. This triggers annoying hourly alerts while the kernel fails to automatically rebalance memory allocation. The issue typically occurs when:

• Applications with memory leaks were previously swapped out
• Kernel vm.swappiness settings are too aggressive
• Long-running processes maintain swapped-out pages unnecessarily

First verify your current memory state:

free -h
              total        used        free      shared  buff/cache   available
Mem:            15G        6.2G        7.1G        456M        2.3G        8.2G
Swap:          2.0G        2.0G         12K

Check which processes are using swap:

sudo smem -t -s swap
PID User     Command                         Swap      USS      PSS      RSS 
4729 mysql    /usr/sbin/mysqld               1.2G    64.3M    65.1M   784.2M
5821 redis    /usr/bin/redis-server         512M     12.1M    12.8M    89.4M

Method 1: Drop Page Cache (Safest approach)

sudo sync; echo 3 | sudo tee /proc/sys/vm/drop_caches

Method 2: Targeted Process Restart

# Identify top swap consumer
sudo swapon --show
sudo swapoff -a && sudo swapon -a

Method 3: Adjust Swappiness Dynamically

sudo sysctl vm.swappiness=10

Create a cron job to prevent future alerts:

#!/bin/bash
SWAP_THRESHOLD=90
CURRENT=$(free | awk '/Swap/{printf "%.0f", $3/$2*100}')

if [ "$CURRENT" -ge "$SWAP_THRESHOLD" ]; then
    echo "Clearing swap at ${CURRENT}% usage"
    sudo swapoff -a && sudo swapon -a
fi

Make it executable and add to cron:

chmod +x /usr/local/bin/swap_monitor.sh
(crontab -l ; echo "*/15 * * * * /usr/local/bin/swap_monitor.sh") | crontab -

For extreme cases where processes won't release swap, use this nuclear option:

# Warning: This will OOM kill processes
echo 1 | sudo tee /proc/sys/vm/oom_kill_allocating_task
sudo sysctl vm.overcommit_memory=2

Consider adding these to /etc/sysctl.conf for permanent changes:

vm.swappiness = 10
vm.vfs_cache_pressure = 50