How to troubleshoot high RAM usage

Quick answer: Check real memory pressure with free -h — look at available, not "free" (Linux intentionally fills RAM with cache). If available is truly low, find the consumers with ps aux --sort=-%mem | head, check the kernel log for OOM kills (dmesg | grep -i oom), add swap as a buffer, and fix or right-size the biggest consumer.

Overview

The most common "high RAM" report is a false alarm: cached memory that Linux returns instantly when needed. Real pressure shows up as low available memory, swap thrashing and the OOM killer terminating processes — typically the database. This guide separates the two and fixes the real case.

Before you start

• SSH access.
• Basic idea of what should run on the server (web server, PHP-FPM, MySQL…).

Step-by-step guide

1. Assess real pressure:

   free -h

   Healthy: available is a decent share of total. Trouble: available near zero and swap heavily used.
2. Find consumers:

   ps aux --sort=-%mem | head -10

3. Check for OOM killer events:

   dmesg -T | grep -i -E "oom|killed process"

   If MySQL/MariaDB gets killed, it will restart with an empty cache and poor performance — a classic small-VPS symptom.
4. Add swap as a safety buffer (if none):

   fallocate -l 2G /swapfile
   chmod 600 /swapfile
   mkswap /swapfile
   swapon /swapfile
   echo '/swapfile none swap sw 0 0' >> /etc/fstab

   Swap prevents hard OOM crashes but is no substitute for RAM — if it's used constantly, you need more memory.
5. Right-size the big consumers: lower PHP-FPM pm.max_children, tune MySQL innodb_buffer_pool_size (typically 50–60% of RAM on a dedicated DB server, much less on a shared box), restart leaking apps on schedule.
6. Still tight? Upgrade RAM — it's the cheapest fix for a database-backed site that has outgrown its plan.

Common issues

• "RAM 95% used" panic: if available is fine, that's cache working as designed — no action needed.
• MySQL keeps dying at night? OOM during backups/cron peaks — add swap, lower buffer sizes, or add RAM.
• Memory grows forever? An app leak — schedule a graceful restart (systemd timer) while you debug.

When to contact support

If memory numbers look impossible for the plan or the server behaves oddly after an upgrade, open a support ticket with free -h and ps output.

Frequently asked questions

Is 90%+ RAM usage always a problem?

No — Linux fills RAM with cache by design. Check the "available" column of free -h; trouble is when available is near zero and swap is thrashing.

What is the OOM killer?

A kernel mechanism that terminates the biggest memory consumer when RAM is exhausted — often the database. Its actions are visible in dmesg.

Does swap replace buying more RAM?

No — swap is a crash safety net. If it is used constantly, performance suffers and you need more RAM.

Related articles

• How to troubleshoot high CPU load
• How to check VPS CPU, RAM and disk usage
• How to upgrade VPS resources
• How to install basic server monitoring

Need a hand? Contact Cloud2Y support →