1numastat(8) Administration numastat(8)
2
6 numastat - Show per-NUMA-node memory statistics for processes and the
7 operating system
8
10 numastat
11 numastat [-V]
13 numastat [<PID>|<pattern>...]
15 numastat [-c] [-m] [-n] [-p <PID>|<pattern>] [-s[<node>]] [-v] [-z]
17 [<PID>|<pattern>...]
18
20 numastat with no command options or arguments at all, displays per-node
21 NUMA hit and miss system statistics from the kernel memory allocator.
22 This default numastat behavior is strictly compatible with the previous
23 long-standing numastat perl script, written by Andi Kleen. The default
24 numastat statistics shows per-node numbers (in units of pages of mem‐
25 ory) in these categories:
26 numa_hit is memory successfully allocated on this node as intended.
28 numa_miss is memory allocated on this node despite the process prefer‐
30 ring some different node. Each numa_miss has a numa_foreign on another
31 node.
32 numa_foreign is memory intended for this node, but actually allocated
34 on some different node. Each numa_foreign has a numa_miss on another
35 node.
36 interleave_hit is interleaved memory successfully allocated on this
38 node as intended.
39 local_node is memory allocated on this node while a process was running
41 on it.
42 other_node is memory allocated on this node while a process was running
44 on some other node.
45 Any supplied options or arguments with the numastat command will sig‐
47 nificantly change both the content and the format of the display.
48 Specified options will cause display units to change to megabytes of
49 memory, and will change other specific behaviors of numastat as
50 described below.
51
53 -c Minimize table display width by dynamically shrinking column
54 widths based on data contents. With this option, amounts of
55 memory will be rounded to the nearest megabyte (rather than the
56 usual display with two decimal places). Column width and inter-
57 column spacing will be somewhat unpredictable with this option,
58 but the more dense display will be very useful on systems with
59 many NUMA nodes.
60 -m Show the meminfo-like system-wide memory usage information.
62 This option produces a per-node breakdown of memory usage infor‐
63 mation similar to that found in /proc/meminfo.
64 -n Show the original numastat statistics info. This will show the
66 same information as the default numastat behavior but the units
67 will be megabytes of memory, and there will be other formatting
68 and layout changes versus the original numastat behavior.
69 -p <PID> or <pattern>
71 Show per-node memory allocation information for the specified
72 PID or pattern. If the -p argument is only digits, it is
73 assumed to be a numerical PID. If the argument characters are
74 not only digits, it is assumed to be a text fragment pattern to
75 search for in process command lines. For example, numastat -p
76 qemu will attempt to find and show information for processes
77 with "qemu" in the command line. Any command line arguments
78 remaining after numastat option flag processing is completed,
79 are assumed to be additional <PID> or <pattern> process speci‐
80 fiers. In this sense, the -p option flag is optional: numastat
81 qemu is equivalent to numastat -p qemu
82 -s[<node>]
84 Sort the table data in descending order before displaying it, so
85 the biggest memory consumers are listed first. With no speci‐
86 fied <node>, the table will be sorted by the total column. If
87 the optional <node> argument is supplied, the data will be
88 sorted by the <node> column. Note that <node> must follow the
89 -s immediately with no intermediate white space (e.g., numastat
90 -s2). Because -s can allow an optional argument, it must always
91 be the last option character in a compound option character
92 string. For example, instead of numastat -msc (which probably
93 will not work as you expect), use numastat -mcs
94 -v Make some reports more verbose. In particular, process informa‐
96 tion for multiple processes will display detailed information
97 for each process. Normally when per-node information for multi‐
98 ple processes is displayed, only the total lines are shown.
99 -V Display numastat version information and exit.
101 -z Skip display of table rows and columns of only zero valuess.
103 This can be used to greatly reduce the amount of uninteresting
104 zero data on systems with many NUMA nodes. Note that when rows
105 or columns of zeros are still displayed with this option, that
106 probably means there is at least one value in the row or column
107 that is actually non-zero, but rounded to zero for display.
108
110 numastat attempts to fold each table display so it will be conveniently
111 readable on the output terminal. Normally a terminal width of 80 char‐
112 acters is assumed. When the resize command is available, numastat
113 attempts to dynamically determine and fine tune the output tty width
114 from resize output. If numastat output is not to a tty, very long out‐
115 put lines can be produced, depending on how many NUMA nodes are
116 present. In all cases, output width can be explicitly specified via
117 the NUMASTAT_WIDTH environment variable. For example, NUMAS‐
118 TAT_WIDTH=100 numastat. On systems with many NUMA nodes, numastat -c
119 -z .... can be very helpful to selectively reduce the amount of dis‐
120 played information.
121
123 NUMASTAT_WIDTH
124
126 /proc/*/numa_maps
127 /sys/devices/system/node/node*/meminfo
128 /sys/devices/system/node/node*/numastat
129
131 numastat -c -z -m -n
132 numastat -czs libvirt kvm qemu
133 watch -n1 numastat
134 watch -n1 --differences=cumulative numastat
135
137 The original numastat perl script was written circa 2003 by Andi Kleen
138 <andi.kleen@intel.com>. The current numastat program was written in
139 2012 by Bill Gray <bgray@redhat.com> to be compatible by default with
140 the original, and to add options to display per-node system memory
141 usage and per-node process memory allocation.
142
144 numactl(8), set_mempolicy(2), numa(3)
145Bill Gray 1.0.0 numastat(8)