The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →To clear swap on a Linux server, first confirm that RAM can hold the swapped pages, then run sudo swapoff -a followed by sudo swapon -a. The first command moves pages from swap back into RAM; the second re-enables every swap device and file configured in /etc/fstab. Check free -h before and after. If available RAM is insufficient, swapoff can stall, fail, or trigger memory exhaustion, so do not run it blindly on a busy production host.
Antihost offers Linux VPS plans with full root access, unlimited traffic on a 1 Gbps port and a 14-day free trial.
Swap clearing is different from dropping the filesystem page cache and different again from changing swappiness. The sections below separate those operations, show when each is appropriate, and provide recovery steps when a server cannot evacuate swap safely.
As an Amazon Associate I earn from qualifying purchases.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →What “clearing swap” means
Linux uses swap as backing storage for memory pages that are not currently resident in physical RAM. A server can show swap in use even when it is operating normally: the kernel may have moved cold pages out of RAM to preserve space for active processes and filesystem cache. A non-zero swap-used value is therefore not, by itself, proof of a fault.
Clearing swap means evacuating those pages into RAM and temporarily disabling the swap areas. The standard sequence is:
#1 Best Overall
sudo swapoff -a
sudo swapon -aswapoff -a applies to all active swap areas. It must find enough usable RAM to receive the pages it is moving. swapon -a then activates the swap partitions and files listed in /etc/fstab. The swap capacity returns, but the “used” amount should be near zero immediately after the migration.
This is an operational reset, not a performance cure. Linux normally reclaims memory and manages swap without manual intervention. If you repeatedly need to clear swap, investigate memory sizing, a leaking process, workload spikes, or an unsuitable storage and swappiness configuration instead of treating the symptom.
Free tools Windows power users keep installed
One-click scans. No signup required.
Check memory headroom before disabling swap
Do these checks in the shell that has administrative privileges:
-
Record the current memory view:
free -hRead the available value in the
Mem:row, not only thefreecolumn. “Available” is the kernel’s estimate of RAM that can be given to applications without swapping heavily. Record theSwap:row as your baseline. -
List every configured swap area and its current use:
swapon --showThis identifies whether swap is a partition, a file, or several areas. It also prevents a surprise when
swapon -alater activates more than one entry.PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteSpecial offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy. -
Look for a process consuming most of the available RAM. Use your normal process monitor, or sort a one-time listing with:
ps -eo pid,comm,%mem,rss --sort=-%mem | headIf available RAM is already low, stop or reschedule a safe workload before attempting the migration. Do not rely on swap as the destination for pages you are trying to remove from swap.
As a practical safety test, the amount of resident RAM that must be available is roughly related to the swap pages currently in use, plus working room for the kernel and applications. The relationship is not exact because pages may be shared, compressed, or reclaimed during the operation. Leave a margin rather than assuming that “available RAM equals swap used” is sufficient.
Clear all swap areas step by step
1. Quiesce memory-heavy work
Pause large builds, database maintenance, backups, or batch jobs if you can. A server that allocates aggressively while pages are being moved has less headroom and is more likely to hit the out-of-memory path.
2. Disable swap and observe the result
sudo swapoff -aThe command may take time while pages are copied back to RAM. In another terminal, watch memory:
watch -n 1 free -hDo not start a second swapoff while the first is still running. If the command returns successfully, the Swap: row should show zero used and zero total until swap is enabled again.
3. Re-enable the configured swap
sudo swapon -aThis reads the swap entries enabled in /etc/fstab. It does not create a new swap area and does not automatically activate an ad-hoc device that is absent from that file. If you intentionally disabled only one area, re-enable that same path explicitly instead:
sudo swapon /path/to/swapfileUse the exact device or file name shown by swapon --show.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors4. Verify capacity and usage
free -h
swapon --showThe swap-used figure should be zero or very small immediately after reactivation. The total should match the areas that were successfully enabled. A missing total usually means an entry in /etc/fstab failed validation, the device is unavailable, or permissions and file setup do not meet the kernel’s requirements.
If swapoff hangs or fails
A stalled operation generally means the system cannot find enough reclaimable RAM to receive all swapped pages. Treat that as a capacity warning, not as a command that should be forced through.
- Check the live figures: run
free -hand inspect the largest processes. If available RAM is falling, stop nonessential work before trying again. - Look for an error message: retain the exact output from
swapoff. It may identify a particular device or an inability to allocate memory. - Reduce demand first: stop a known memory-heavy service using that service’s normal, reversible procedure. Avoid killing an unknown process merely to make the command complete.
- Do not reboot as a first response: rebooting discards the diagnostic state and can turn a controlled memory shortage into an outage. Plan a maintenance window if the host cannot maintain headroom.
- Escalate the underlying issue: repeated failures indicate that the server needs more RAM, a lower-memory workload, or investigation of a leak. Clearing swap is not a substitute for correcting that condition.
If swapoff -a succeeds but swapon -a reports an error, leave the host running while you inspect the specific swap entry. Confirm that the device exists, the file path is correct, and the entry in /etc/fstab has the intended type and options. Re-enable a known-good area explicitly only after confirming the path.
Clear one swap area instead of everything
On a server with multiple swap devices or files, you may need to evacuate only one area for maintenance. First identify the exact target:
Rank #4
swapon --showThen disable and re-enable that path:
sudo swapoff /dev/your-swap-device
sudo swapon /dev/your-swap-deviceFor a swap file, substitute its absolute path. This still requires enough RAM for the pages currently residing on that area. A device-by-device approach reduces the scope of the change but does not remove the headroom requirement.
Confirm that you cleared swap, not the page cache
Swap and filesystem cache are different memory consumers. The swapoff/swapon sequence moves swapped anonymous pages into RAM. It does not deliberately remove clean file data from the page cache.
If your goal is to test application behavior with a cold cache, use the kernel’s cache-control interface instead. Flush pending writes first:
syncThen choose the level that matches the test:
| Command | Objects affected | Use |
|---|---|---|
echo 1 | sudo tee /proc/sys/vm/drop_caches |
Clean page cache | Remove cached file contents |
echo 2 | sudo tee /proc/sys/vm/drop_caches |
Reclaimable slab objects, including dentries and inodes | Remove filesystem metadata caches |
echo 3 | sudo tee /proc/sys/vm/drop_caches |
Both page cache and reclaimable slab objects | Clear both categories for a controlled test |
sync matters because dirty objects cannot be freed by drop_caches until their data is written. Dropping cache does not cap future cache growth; Linux will build the cache again as files are read. Use it manually for diagnostics or benchmarking, not as a routine memory-management job.
Kernel messages emitted when caches are dropped are informational. The control value uses bits 1 and 2 for the two cache classes; writing value 4 (bit 2) suppresses those messages, but it does not itself free cache. Do not confuse that logging bit with a cache-clearing level.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Use swappiness to change future behavior
vm.swappiness is a policy setting, not a clearing command. It expresses the relative I/O cost of swapping versus paging filesystem data on a scale from 0 to 200. The documented default is 60. A value of 100 treats the two costs as equal; lower values make swap appear more expensive, so the kernel tends to preserve anonymous pages in RAM, while higher values make swap appear cheaper.
Best Value
| Setting direction | Kernel preference | When it may fit |
|---|---|---|
| Lower than 60 | Prefer keeping application pages in RAM and reclaim filesystem cache sooner | Workloads where latency from swapped application pages is especially harmful |
| 60 | Default balance | Starting point when you have no workload evidence for a change |
| 100 | Treat swapping and filesystem paging as equal-cost choices | Systems where those I/O paths have similar costs |
| Above 100 | Make swap look relatively inexpensive | Some in-memory swap or hybrid setups, such as zram, zswap, or swap on storage faster than the filesystem workload |
Inspect and change the live value with:
sysctl vm.swappiness
sudo sysctl -w vm.swappiness=20The second command lasts until the next reboot. To persist a chosen value, create a file such as /etc/sysctl.d/99-swappiness.conf containing:
vm.swappiness = 20Then load the configuration:
sudo sysctl --systemChoose a value through measurement and workload testing, not by copying a generic “best” number. Memory cgroups can apply a group-specific policy that overrides the global setting, so a container or service may behave differently from the host.
zram, zswap, and compressed swap considerations
Not all swap is a disk partition or ordinary swap file. Some distributions use compressed RAM devices such as zram, and systems may use zswap as a compressed front end. On those configurations, disabling swap can require decompressing pages back into regular RAM. The resulting RAM spike can be substantial even when the swap device itself is small.
Identify the implementation before a production change:
swapon --show
cat /proc/swapsIf the type or device name indicates compressed swap, test the exact procedure on a non-production host with a similar memory limit. Keep a console or out-of-band access path available in case the host becomes unresponsive. The ordinary swapoff -a/swapon -a sequence still describes the operation, but its memory impact is different from evacuating pages from a disk-backed area.
When clearing swap is appropriate
- Planned storage maintenance: evacuate a swap device before replacing, resizing, or detaching it, provided the host has sufficient RAM.
- Controlled diagnostics: establish a known starting point for a memory or latency test, then record the workload and cache state so results are comparable.
- Post-incident cleanup: remove stale swapped pages after a temporary memory spike, once the host has ample headroom.
- Routine monitoring alerts: do not clear swap simply because a graph is non-zero. Examine pressure, latency, available RAM, and process behavior first.
If a server is under sustained memory pressure, adding physical RAM, reducing concurrency, moving a workload, or fixing a leak is usually safer than repeatedly cycling swap. A swap reset can make the graph look clean while leaving the cause untouched.
Quick Recap
Operational checklist
- Schedule the change and ensure console or out-of-band access.
- Run
free -hand record available RAM and swap used. - Run
swapon --showto enumerate every swap area. - Pause avoidable memory-heavy work.
- Run
sudo swapoff -aand wait for it to complete. - Run
sudo swapon -ato restore configured capacity. - Verify with
free -handswapon --show. - Investigate any failure before repeating the operation.
- Use
syncanddrop_cachesonly when the objective is page-cache testing, not swap clearing. - Tune
vm.swappinessonly after observing the workload, and document any persistentsysctl.dchange.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
