Linux caches as much data as possible in RAM to speed up disk reads. When applications need more memory, the kernel automatically reclaims this cache. However, in some situations — benchmarking, after bulk file operations, or troubleshooting — you may want to manually drop the page cache, dentries, and inodes. This guide explains how, and when you should (and shouldn’t) do it.

Understanding Linux Memory Cache

Running free -h shows a “buff/cache” column that can look alarming. This memory is not wasted — it’s used productively to cache files and will be released automatically when applications need RAM. You rarely need to clear it manually on a production server.

free -h
#               total    used    free   shared  buff/cache  available
# Mem:           16Gi    4.2Gi    892Mi   312Mi    10.8Gi     11.2Gi

How to Drop Linux Page Cache

# Step 1: Flush dirty pages to disk first (always do this before dropping cache)
sync

# Step 2: Choose what to drop:
# Drop page cache only
echo 1 > /proc/sys/vm/drop_caches

# Drop dentries and inodes only
echo 2 > /proc/sys/vm/drop_caches

# Drop page cache, dentries, and inodes (most thorough)
echo 3 > /proc/sys/vm/drop_caches

Combine both steps in one line:

sync && echo 3 > /proc/sys/vm/drop_caches

Verify the Cache Was Cleared

free -h
# The buff/cache value should now be significantly lower

Permission Denied on VPS / Containers

On OpenVZ/Virtuozzo containers, /proc/sys/vm/drop_caches is read-only from inside the container. Cache management must be done at the hardware node level by your hosting provider. This is expected behaviour — contact your host’s support team.

When NOT to Clear the Cache

• Never on production under load — dropping the cache causes a flood of disk I/O as the kernel rebuilds it, which can spike load average significantly
• Not as a regular maintenance task — the kernel manages cache automatically and efficiently
• Not to “free up memory” for applications — the kernel already does this automatically when apps need RAM