Skip to content

vmstat

Shows in one line per interval how busy CPU, memory, swap and disks are; the quickest overview when a server is slow.

vmstat output

Syntax

vmstat [OPTIONS] [INTERVAL [COUNT]]

Options

Option Description
INTERVAL Print a new line every N seconds
COUNT Stop after N lines
-w Wide output, columns don't run into each other
-S M Show memory in MB instead of KB
-s Memory and event counters as a list
-d Statistics per disk
-t Add a timestamp to every line

The most important columns:

Column Meaning Watch out when
r Processes waiting for a CPU permanently higher than the number of cores
b Processes blocked, usually waiting for disk I/O not 0 for a longer time
si, so Swapped in and out per second not 0: memory is short
us, sy CPU time for programs and kernel together near 100
wa CPU waits for I/O above about 10: disk is the bottleneck
st Time stolen by the hypervisor not 0 on a VM: the host is overloaded

Examples

Five measurements, every two seconds:

vmstat -S M 2 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    812     98   1798    0    0    14    38  112  201  3  1 96  0  0  0
 0  0      0    811     98   1798    0    0     0    24  148  263  2  1 97  0  0  0
 1  0      0    811     98   1798    0    0     0     6  139  241  2  0 98  0  0  0
 0  0      0    810     98   1798    0    0     0    40  162  287  3  1 96  0  0  0
 0  0      0    810     98   1798    0    0     0     0  127  230  1  0 99  0  0  0

An idle server: id (idle) near 100, no swap activity, no waiting processes.

A server that waits for the disk:

vmstat -S M 2 3
procs -----------memory---------- ---swap-- -----io---- -system-- -------cpu-------
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st gu
 1  3      0    205     98   2310    0    0  8420 21650 1890 2410  6  9 37 48  0  0
 0  4      0    198     98   2321    0    0  9104 23102 1977 2533  5 10 36 49  0  0
 1  3      0    201     98   2319    0    0  8833 22410 1920 2487  6  9 35 50  0  0

b is 3 to 4 and wa around 50 %: processes wait for the disk while the CPU has time. Next step: iostat -x and iotop.

Memory is short and the server swaps:

vmstat -S M 2 3
procs -----------memory---------- ---swap-- -----io---- -system-- -------cpu-------
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st gu
 3  1    612     54      4    121  310  498  1310  1902 1488 1720 41 12 22 25  0  0
 2  2    640     49      4    118  352  514  1402  2011 1530 1798 39 14 21 26  0  0
 4  1    671     51      3    115  288  476  1250  1876 1501 1755 43 12 20 25  0  0

si and so are permanently above zero: the server constantly moves memory to disk and back. Find the process that needs so much memory, see top.

When do I need it?

  • As the first command when a server is slow: CPU, memory or disk?
  • To watch a load test or a backup run live
  • To find out whether a VM suffers from an overloaded host (st)

Ignore the first line

The first line shows averages since the boot, not the current state. Look at the following lines, such as with vmstat 2 5.

iostat · sar · top · free