How to check memory pressure in Linux

Memory pressure describes how often Linux pauses work to reclaim memory, which typically shows up as latency spikes long before an out-of-memory event.

Reclaim happens when the kernel needs free pages for new allocations, dropping page cache first and swapping anonymous memory when necessary. Pressure Stall Information (PSI) exposes reclaim-induced stalls via /proc/pressure/memory, providing time-windowed averages that track real slowdowns more directly than a single “free RAM” snapshot.

Short bursts are normal during cache churn, builds, or restarts. Sustained non-zero PSI averages combined with swap-in activity often indicates RAM contention, an oversized workload, or an enforced memory limit in a cgroup.

1. Review overall memory availability plus swap status.

   $ free -h
                  total        used        free      shared  buff/cache   available
   Mem:            11Gi       1.2Gi       8.0Gi        48Mi       2.8Gi        10Gi
   Swap:          4.0Gi          0B       4.0Gi

   available reflects reclaimable cache; consistently low available is more meaningful than low free.

2. Sample swap-in plus swap-out rates over a short interval.

   $ vmstat 1 5
   procs -----------memory---------- ---swap-- -----io---- -system-- -------cpu-------
    r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st gu
    0  0      0 8396360 284400 2621276    0    0   107   267  551    0  0  0 99  0  0  0
    0  0      0 8440064 284400 2621276    0    0     0     0  386  494  0  0 100  0  0  0
    0  0      0 8440064 284400 2621276    0    0     0     0  422  535  0  0 100  0  0  0
    0  0      0 8440064 284400 2621276    0    0     0     0  190  223  0  0 100  0  0  0
    0  0      0 8440064 284404 2621276    0    0     0    20  263  311  0  0 100  0  0  0

   The first report line is an average since boot; focus on the interval lines that follow.

   Sustained non-zero si indicates swap-in, which commonly correlates with reclaim stalls.

3. Read memory Pressure Stall Information metrics.

   $ cat /proc/pressure/memory
   some avg10=0.00 avg60=0.00 avg300=0.00 total=1451483
   full avg10=0.00 avg60=0.00 avg300=0.00 total=1445159

   some tracks time when at least one task is stalled on reclaim; full tracks time when all runnable tasks are stalled.

   avg10, avg60, and avg300 represent the percentage of wall time spent stalled in each window; total is cumulative stall time in microseconds.

4. Watch PSI averages during an observed slowdown.

   $ timeout 2 cat /proc/pressure/memory
   some avg10=0.00 avg60=0.00 avg300=0.00 total=1451483
   full avg10=0.00 avg60=0.00 avg300=0.00 total=1445159

   Non-zero full pressure indicates system-wide stalls; interactive latency typically becomes obvious at that point.

5. Locate cgroup v2 memory pressure files when troubleshooting containers or systemd services.

   $ find /sys/fs/cgroup/system.slice -maxdepth 3 -name memory.pressure | head
   find: ‘/sys/fs/cgroup/system.slice’: No such file or directory

   Read any listed memory.pressure file the same way as /proc/pressure/memory for a per-cgroup view.

6. Confirm pressure returns near zero during idle periods.

   $ cat /proc/pressure/memory
   some avg10=0.00 avg60=0.00 avg300=0.00 total=1451483
   full avg10=0.00 avg60=0.00 avg300=0.00 total=1445159