How to Diagnose and Fix High Swap Usage on Linux Servers

Even when memory appears available, a server can become sluggish, time out, or have processes killed because swap usage is extremely high.

Step 1: Check Swap Usage

View overall Swap usage

free -h
# Example output
               total   used   free  shared  buff/cache   available
Mem:           15Gi   10Gi   1.2Gi   200Mi        4.0Gi        4.8Gi
Swap:          8.0Gi   7.5Gi   500Mi

Check swap activity (critical): vmstat 1 5 Focus on the last two columns: si (swap‑in) and so (swap‑out). Normal: both stay at 0. Abnormal: persistently >0 indicates frequent paging.

Step 2: Find the Process Consuming Swap

Method 1: Sort processes by swap usage (recommended)

for file in /proc/*/status; do
  awk '/VmSwap|Name/{printf $2 " " $3}END{ print ""}' $file;
done | sort -k 2 -n -r | head -10

Example output shows java 2145632 kB, meaning the Java process is using about 2.1 GB of swap.

Method 2: Use smem tool (more visual)

# Install
yum install -y smem      # CentOS/Rocky
apt install -y smem      # Ubuntu/Debian

# Sort by swap
smem -s swap -r | head -10

Step 3: Targeted Fixes

Case 1: Process memory leak (e.g., Java, Python)

Restart service (temporary relief)
Analyze memory dump (e.g., jmap -dump for Java)
Limit maximum memory (e.g., JVM -Xmx4g)

Case 2: System memory sufficient but swap still used

Reason: Linux’s default swappiness (60) makes it use swap aggressively.

# Temporary change
sysctl vm.swappiness=10

# Permanent change
echo 'vm.swappiness=10' >> /etc/sysctl.conf

Suggested values: databases/high‑performance services 1‑10, normal servers 30, never use swap 0 (only if memory is ample).

Case 3: Large cache triggers swap

# Release pagecache (safe)
echo 1 > /proc/sys/vm/drop_caches
# Release dentries and inodes (moderately safe)
echo 2 > /proc/sys/vm/drop_caches
# Release all (use with caution in production)
echo 3 > /proc/sys/vm/drop_caches

⚠️ This only provides temporary relief; the root cause remains swappiness or application memory management.

Swap is not the enemy, but frequent use is a performance disaster!

Automation script

#!/bin/bash
echo "=== Swap usage overview ==="
free -h

echo -e "
=== Swap activity (si/so) ==="
vmstat 1 3 | tail -1

echo -e "
=== Top 10 swap‑using processes ==="
for file in /proc/*/status; do
  awk '/VmSwap|Name/{printf $2 " " $3}END{ print ""}' $file 2>/dev/null;
done | sort -k 2 -n -r | head -10

echo -e "
=== Current swappiness ==="
cat /proc/sys/vm/swappiness

chmod +x swap-check.sh
./swap-check.sh

Regularly monitor your server’s state and adjust configurations to avoid performance bottlenecks.