1PMLOGREDUCE(1) General Commands Manual PMLOGREDUCE(1)
2
6 pmlogreduce - temporal reduction of Performance Co-Pilot archives
7
9 $PCP_BINADM_DIR/pmlogreduce [-z] [-A align] [-S starttime] [-s samples]
10 [-T endtime] [-t interval] [-v volsamples] [-Z timezone] input output
11
13 pmlogreduce reads one Performance Co-Pilot (PCP) archive identified by
14 input (this must be a PCP archive created by pmlogger(1), pmlogex‐
15 tract(1) or pmlogreduce(1)), and creates a temporally reduced PCP ar‐
16 chive in output. The data reduction involves statistical and temporal
17 reduction of samples with an output sampling interval defined by the -t
18 option in the output archive (independent of the sampling intervals in
19 the input archive), and is further controlled by other command line
20 arguments.
21 For some metrics, temporal data reduction is not going to be helpful,
23 so for metrics with types PM_TYPE_AGGREGATE or PM_TYPE_EVENT, a warning
24 is issued if these metrics are found in input and they will be skipped
25 and not appear in the output archive.
26
28 The command line options for pmlogreduce are as follows:
29 -A align
31 Specify a ``natural'' alignment of the output sample times;
32 refer to PCPIntro(1).
33 -S starttime
35 Define the start of a time window to restrict the samples
36 retrieved from the input archive; refer to PCPIntro(1).
37 -s samples
39 The argument samples defines the number of samples to be written
40 to output. If samples is 0 or -s is not specified, pmlogreduce
41 will sample until the end of the PCP archive, or the end of the
42 time window as specified by -T, whichever comes first. The -s
43 option will override the -T option if it occurs sooner.
44 -T endtime
46 Define the termination of a time window to restrict the samples
47 retrieved from the input archive; refer to PCPIntro(1).
48 -v volsamples
50 The output archive is potentially a multi-volume data set, and
51 the -v option causes pmlogreduce to start a new volume after
52 volsamples log records have been written to the output archive.
53 Independent of any -v option, each volume of an archive is lim‐
55 ited to no more than 2^31 bytes, so pmlogreduce will automati‐
56 cally create a new volume for the archive before this limit is
57 reached.
58 -t interval
60 Consecutive samples in the output archive will appear with a
61 time delta defined by interval; refer to PCPIntro(1). Note the
62 default value is 600 (seconds, i.e. 10 minutes).
63 -Z timezone
65 Use timezone when displaying the date and time, or interpreting
66 the -S and -T options. Timezone is in the format of the envi‐
67 ronment variable TZ as described in environ(7).
68 -z Use the local timezone of the host from the input archive when
70 displaying the date and time, or interpreting the -S and -T
71 options. The default is to initially use the timezone of the
72 local host.
73
75 The statistical and temporal reduction follows the following rules:
76 1. Consecutive records from input are read without interpolation, and
78 at most one output record is written for each interval, summarizing
79 the performance data over that period.
80 2. If the semantics of a metric indicates it is instantaneous or dis‐
82 crete then output value is computed as the arithmetic mean of the
83 observations (if any) over each interval.
84 3. If the semantics of a metric indicates it is a counter then the
86 following transformations are applied:
87 a) Metrics with 32-bit precision are promoted to 64-bit precision.
88 b) Any counter wrap (overflow) is noted, and appropriate adjust‐
89 ment made in the value of the metric over each interval. This
90 will be correct in the case of a single counter wrap, but will
91 silently underestimate in the case where more than one counter
92 wrap occurs between consecutive observations in the input ar‐
93 chive, and silently overestimate in the case where a counter is
94 reset occurs between consecutive observations in the input ar‐
95 chive; unfortunately these situations cannot be detected, but
96 are believed to be rare events for the sort of production moni‐
97 toring environments where pmlogreduce is most likely to be
98 deployed.
99 4. Any changes in instance domains, and indeed all metadata, is pre‐
101 served.
102 5. Any ``mark'' records in the input archive (as created by pmlogex‐
104 tract(1)) will be preserved in the output archive, so periods where
105 no data is available are maintained, and data interpolation will
106 not occur across these periods when the output archive is subse‐
107 quently processed with PCP applications.
108
110 For each of the input and output archives, several physical files are
111 used.
112 archive.meta
113 metadata (metric descriptions, instance domains, etc.) for
114 the archive log
115 archive.0 initial volume of metrics values (subsequent volumes have
116 suffixes 1, 2, ...) - for input these files may have been
117 previously compressed with bzip2(1) or gzip(1) and thus may
118 have an additional .bz2 or .gz suffix.
119 archive.index
120 temporal index to support rapid random access to the other
121 files in the archive log.
122
124 Environment variables with the prefix PCP_ are used to parameterize the
125 file and directory names used by PCP. On each installation, the file
126 /etc/pcp.conf contains the local values for these variables. The
127 $PCP_CONF variable may be used to specify an alternative configuration
128 file, as described in pcp.conf(5).
129
131 PCPIntro(1), pmdumplog(1), pmlc(1), pmlogextract(1), pmlogger(1),
132 pcp.conf(5) and pcp.env(5).
133
135 All error conditions detected by pmlogreduce are reported on stderr
136 with textual (if sometimes terse) explanation.
137 Should the input archive be corrupted (this can happen if the pmlogger
139 instance writing the archive suddenly dies), then pmlogreduce will
140 detect and report the position of the corruption in the file, and any
141 subsequent information from the input archive will not be processed.
142 If any error is detected, pmlogreduce will exit with a non-zero status.
144
146 The preamble metrics (pmcd.pmlogger.archive, pmcd.pmlogger.host, and
147 pmcd.pmlogger.port), which are automatically recorded by pmlogger at
148 the start of the archive, may not be present in the archive output by
149 pmlogreduce. These metrics are only relevant while the archive is
150 being created, and have no significance once recording has finished.
151Performance Co-Pilot PCP PMLOGREDUCE(1)