1PCP-COLLECTL(1) General Commands Manual PCP-COLLECTL(1)
2
6 pmcollectl, pcp-collectl - collect data that describes the current sys‐
7 tem status
8
10 pcp collectl [-f file | -p file ...] [options ...]
11
13 pcp-collectl is a system-level performance monitoring utility that
14 records or displays specific operating system data for one or more sets
15 of subsystems. Any of the subsystems (such as CPU, Disks, Memory or
16 Sockets) can be included or excluded from data collection. Data can
17 either be displayed immediately to a terminal, or stored in files for
18 retrospective analysis.
19 pcp-collectl is a python(1) script providing much of the functionality
21 available from the collectl(1) Linux utility (which happens to be writ‐
22 ten in perl(1)).
23 It makes use of the Performance Co-Pilot (PCP) toolkit to simplify its
25 implementation, as well as provide more of the collectl functionality
26 on platforms other than Linux.
27 pcp-collectl has two primary modes of operation:
29 1. Record Mode (-f or --filename option) which reads data from a live
31 system and writes output to a file or displays it on a terminal.
32 2. Playback Mode (-p or -a option) which reads data from one or more
34 PCP archive files and displays output on a terminal. Note that
35 these files are not raw collectl format data, rather they are ar‐
36 chives created by the pmlogger(1) utility (possibly indirectly,
37 through use of the -f option to pcp-collectl).
38
40 In this mode data is taken from a live system and either displayed on
41 the terminal or written to a PCP archive.
42 -h host
44 Display metrics from host instead of displaying metrics from the
45 local host.
46 -c, --count samples
48 The number of samples to record.
49 -f, --filename filename
51 This is the name of a PCP archive to write the output to.
52 -i, --interval interval
54 This is the sampling interval in seconds. The default is 1 sec‐
55 ond.
56 -R, --runtime duration
57 Specify the duration of data collection where the duration is a
58 number followed by one of wdhms, indicating how many weeks,
59 days, hours, minutes or seconds the collection is to be taken
60 for.
61
63 In this mode, data is read from one or more PCP data files that were
64 generated with the recording option, or indirectly via the pmlogger
65 utility.
66 -f, --filename filename
68 If specified, this is the name of a PCP archive to write the
69 output to (rather than the terminal).
70 -p, --playback filename
72 Read data from the specified PCP archive folio files(s) - refer
73 to pmafm(1) for archive folio details.
74 -a, --archive filename
76 Read data from the specified PCP raw archive files(s).
77
79 The following options are supported in both record and playback modes.
80 --help
82 Display standard help message.
83 -s, --subsys subsystem
85 This field controls which subsystem data is to be collected or
86 played back for. The rules for displaying results vary depending
87 on the type of data to be displayed. If you write data for CPUs
88 and DISKs to a raw file and play it back with -sc, you will only
89 see CPU data. If you play it back with -scm you will still only
90 see CPU data since memory data was not collected. To see the
91 current set of default subsystems, which are a subset of this
92 full list, use -h.
93 The default is "cdn", which stands for CPU, Disk and Network
95 summary data.
96 SUMMARY SUBSYSTEMS
98 c - CPU
100 d - Disk
101 f - NFS V3 Data
102 j - Interrupts
103 m - Memory
104 n - Networks
105 y - Slabs (system object caches)
106 DETAIL SUBSYSTEMS
108 This is the set of detail data from which in most cases the cor‐
110 responding summary data is derived. So, if one has 3 disks and
111 chooses -sd, one will only see a single total taken across all 3
112 disks. If one chooses -sD, individual disk totals will be
113 reported but no totals.
114 C - CPU
116 D - Disk
117 F - NFS Data
118 J - Interrupts
119 M - Memory node data, which is also known as NUMA data
120 N - Networks
121 Y - Slabs (system object caches)
122 Z - Processes
123 --verbose
125 Display output in verbose mode. This often displays more data
126 than in the default mode. When displaying detail data, verbose
127 mode is forced. Furthermore, if summary data for a single sub‐
128 system is to be displayed in verbose mode, the headers are only
129 repeated occasionally whereas if multiple subsystems are
130 involved each needs their own header.
131
133 PCPIntro(1), collectl(1), collectl2pcp(1), perl(1), python(1), pmlog‐
134 ger(1), pmcd(1), pmafm(1), pmprobe(1), pmrep(1), PMAPI(3), and
135 pcp.conf(5).
136Performance Co-Pilot PCP PCP-COLLECTL(1)