Tuning vm.* Kernel Parameters: Swappiness, Dirty Ratio, and OOM Behavior on a VPS

vm.vfs_cache_pressure: Keeping Inode/Dentry Caches Under Control

The VFS (Virtual File System) cache stores recently accessed directory entries (dentries) and inodes. On a busy VPS with many files (e.g., a WordPress site with thousands of cached pages), this cache can grow large. The vfs_cache_pressure parameter controls how aggressively the kernel reclaims these caches.

• Default (100): Balanced reclamation
• Lower (50): Keep cache entries longer — more memory used, better performance for repeated file access
• Higher (200): Reclaim cache aggressively — frees memory at the cost of more disk I/O

For a web server VPS, setting vfs_cache_pressure to 50 can improve performance by keeping frequently accessed file metadata in memory:

sudo sysctl -w vm.vfs_cache_pressure=50
sudo sh -c 'echo "vm.vfs_cache_pressure=50" >> /etc/sysctl.conf'

OOM Behavior: vm.overcommit_memory and vm.overcommit_ratio

Linux overcommits memory by default: it allows processes to allocate more virtual memory than physical RAM + swap. When a process actually tries to use the memory, the kernel must find pages to free. If it cannot, the OOM (Out-Of-Memory) killer terminates a process.

On a VPS, the default overcommit behavior can lead to unpredictable OOM kills at the worst possible moment. The safest setting is vm.overcommit_memory=2, which disables overcommit and allocates memory strictly based on available RAM + swap:

# Disable overcommit (strict accounting)
sudo sysctl -w vm.overcommit_memory=2
sudo sysctl -w vm.overcommit_ratio=50  # Allow 50% overhead for allocations

# Make permanent
sudo sh -c 'echo "vm.overcommit_memory=2" >> /etc/sysctl.conf'
sudo sh -c 'echo "vm.overcommit_ratio=50" >> /etc/sysctl.conf'

With overcommit_memory=2, the kernel will reject memory allocations that exceed available capacity. Applications that try to allocate more memory than exists will fail immediately with an error instead of causing a random OOM kill later. This is much easier to debug.

vm.min_free_kbytes: Reserving Memory for Critical Allocations

This parameter reserves a minimum amount of free memory for critical kernel allocations. If the system runs out of memory entirely, the kernel cannot even run the OOM killer. Setting a reasonable reserve prevents this deadlock scenario.

# Reserve 64 MB on a 2 GB VPS (roughly 3%)
sudo sysctl -w vm.min_free_kbytes=65536
sudo sh -c 'echo "vm.min_free_kbytes=65536" >> /etc/sysctl.conf'

A good rule of thumb is 3-5% of total RAM. On a 1 GB VPS, set min_free_kbytes to 32768 (32 MB). On a 4 GB VPS, set it to 131072 (128 MB).

Complete vm.* sysctl Configuration for a VPS

Here is a complete set of vm.* parameters tuned for a low-to-mid-range VPS (1-4 GB RAM, SSD storage):

# /etc/sysctl.d/99-vps-vm.conf
# Memory management tuning for VPS

# Reduce swap tendency
vm.swappiness=10

# Flush dirty pages more frequently
vm.dirty_background_ratio=5
vm.dirty_ratio=10

# Keep directory cache in memory longer
vm.vfs_cache_pressure=50

# Strict overcommit policy
vm.overcommit_memory=2
vm.overcommit_ratio=50

# Reserve memory for critical kernel allocations
vm.min_free_kbytes=65536

# Reduce page-cluster size (pages read ahead on swap)
vm.page-cluster=0

Apply with sudo sysctl -p /etc/sysctl.d/99-vps-vm.conf. Monitor memory usage with free -h and vmstat 1 after applying to verify the system behaves as expected.

Monitoring and Validation

After tuning, validate that your changes are having the desired effect:

# Check current values
sysctl vm.swappiness vm.dirty_ratio vm.dirty_background_ratio

# Monitor swap usage
free -h
swapon --show

# Check OOM status
dmesg | grep -i "oom"

# Monitor page faults and swap activity
vmstat 1 10

If you see swap usage increasing after tuning, your swappiness value may still be too high, or your VPS simply needs more RAM. For persistent memory pressure, compare VPS plans on our homepage to find a configuration with adequate memory for your workload.

Conclusion

The vm.* kernel parameters directly control how your VPS handles memory pressure, disk I/O, and process stability. By tuning swappiness, dirty page ratios, overcommit behavior, and cache pressure, you can reduce latency, prevent OOM kills, and make more efficient use of limited RAM. Apply these settings on every new VPS as part of your initial hardening process.

Related Articles：

Memory Budgeting for a Small VPS: How to Fit Nginx, PHP-FPM, and MySQL in 1–2 GB of RAM VPS Swap Space Configuration: When and How to Use Swap for Better Performance Monitoring VPS Resource Usage: Netdata vs Prometheus – A Practical Comparison Reading VPS Benchmark Results: How to Interpret CPU Steal, Load Average, and iowait