1PMLOGREDUCE(1)              General Commands Manual             PMLOGREDUCE(1)
2
3
4

NAME

6       pmlogreduce - temporal reduction of Performance Co-Pilot archives
7

SYNOPSIS

9       pmlogreduce [-z?]  [-A align] [-s samples] [-S starttime] [-t interval]
10       [-T endtime] [-v volsamples] [-Z timezone] input output
11

DESCRIPTION

13       pmlogreduce reads one set of Performance Co-Pilot (PCP) archives  iden‐
14       tified by input and creates a temporally reduced PCP archive in output.
15       input is a comma-separated list of names, each of which may be the base
16       name  of  an  archive or the name of a directory containing one or more
17       archives.  The data reduction involves statistical and temporal  reduc‐
18       tion  of samples with an output sampling interval defined by the -t op‐
19       tion in the output archive (independent of the  sampling  intervals  in
20       the  input  archives),  and is further controlled by other command line
21       arguments.
22
23       For some metrics, temporal data reduction is not going to  be  helpful,
24       so for metrics with types PM_TYPE_AGGREGATE or PM_TYPE_EVENT, a warning
25       is issued if these metrics are found in input and they will be  skipped
26       and not appear in the output archive.
27

OPTIONS

29       The available command line options are:
30
31       -A align, --align=align
32            Specify  a ``natural'' alignment of the output sample times; refer
33            to PCPIntro(1).
34
35       -s samples, --samples=samples
36            The argument samples defines the number of samples to  be  written
37            to  output.   If  samples is 0 or -s is not specified, pmlogreduce
38            will sample until the end of the set of PCP archives, or  the  end
39            of the time window as specified by -T, whichever comes first.  The
40            -s option will override the -T option if it occurs sooner.
41
42       -S starttime, --start=starttime
43            Define the start of a time window  to  restrict  the  samples  re‐
44            trieved from the input archives; refer to PCPIntro(1).
45
46       -t interval, --interval=interval
47            Consecutive  samples in the output archive will appear with a time
48            delta defined by interval; refer to PCPIntro(1).  Note the default
49            value is 600 (seconds, i.e. 10 minutes).
50
51       -T endtime, --finish=endtime
52            Define  the  termination  of a time window to restrict the samples
53            retrieved from the input archives; refer to PCPIntro(1).
54
55       -v volsamples
56            The output archive is potentially a multi-volume data set, and the
57            -v  option  causes pmlogreduce to start a new volume after volsam‐
58            ples records have been written to the output archive.
59
60              Independent of any -v option, each volume of an archive is  lim‐
61              ited  to  no more than 2^31 bytes, so pmlogreduce will automati‐
62              cally create a new volume for the archive before this  limit  is
63              reached.
64
65       -z, --hostzone
66            Use  the  local  timezone of the host from the input archives when
67            displaying the date and time, or interpreting the -S  and  -T  op‐
68            tions.   The default is to initially use the timezone of the local
69            host.
70
71       -Z timezone, --timezone=timezone
72            Use timezone when displaying the date and  time,  or  interpreting
73            the  -S and -T options.  Timezone is in the format of the environ‐
74            ment variable TZ as described in environ(7).
75
76       -?, --help
77            Display usage message and exit.
78

DATA REDUCTION

80       The statistical and temporal reduction follows the following rules:
81
82       1.  Consecutive records from input are read without interpolation,  and
83           at most one output record is written for each interval, summarizing
84           the performance data over that period.
85
86       2.  If the semantics of a metric indicates it is instantaneous or  dis‐
87           crete  then  output value is computed as the arithmetic mean of the
88           observations (if any) over each interval.
89
90       3.  If the semantics of a metric indicates it is  a  counter  then  the
91           following transformations are applied:
92           a)  Metrics with 32-bit precision are promoted to 64-bit precision.
93           b)  Any  counter  wrap (overflow) is noted, and appropriate adjust‐
94               ment made in the value of the metric over each interval.   This
95               will  be correct in the case of a single counter wrap, but will
96               silently underestimate in the case where more than one  counter
97               wrap  occurs  between consecutive observations in the input ar‐
98               chives, and silently overestimate in the case where  a  counter
99               reset  occurs between consecutive observations in the input ar‐
100               chives; unfortunately these situations cannot be detected,  but
101               are believed to be rare events for the sort of production moni‐
102               toring environments where pmlogreduce is most likely to be  de‐
103               ployed.
104
105       4.  Any  changes  in instance domains, and indeed all metadata, is pre‐
106           served.
107
108       5.  Any ``mark'' records in the input archives (as created by  pmlogex‐
109           tract(1)) will be preserved in the output archive, so periods where
110           no data is available are maintained, and  data  interpolation  will
111           not  occur  across  these periods when the output archive is subse‐
112           quently processed with PCP applications.
113

CAVEATS

115       The preamble metrics  (pmcd.pmlogger.archive,  pmcd.pmlogger.host,  and
116       pmcd.pmlogger.port),  which  are  automatically recorded by pmlogger at
117       the start of the archive, may not be present in the archive  output  by
118       pmlogreduce.   These metrics are only relevant while the archive is be‐
119       ing created, and have no significance once recording has finished.
120

DIAGNOSTICS

122       All error conditions detected by pmlogreduce  are  reported  on  stderr
123       with textual (if sometimes terse) explanation.
124
125       Should the input archives be corrupted (this can happen if the pmlogger
126       instance writing the archive suddenly dies), then pmlogreduce will  de‐
127       tect  and  report  the  position of the corruption in the file, and any
128       subsequent information from the input archives will not be processed.
129
130       If any error is detected, pmlogreduce will exit with a non-zero status.
131

FILES

133       For each of the input and output archives, several physical  files  are
134       used.
135
136       archive.meta
137            metadata (metric descriptions, instance domains, etc.) for the ar‐
138            chive
139
140       archive.0
141            initial volume of metrics values (subsequent volumes have suffixes
142            1,  2,  ...) - for input these files may have been previously com‐
143            pressed with bzip2(1) or gzip(1) and thus may have  an  additional
144            .bz2 or .gz suffix.
145
146       archive.index
147            temporal  index  to support rapid random access to the other files
148            in the archive.
149

PCP ENVIRONMENT

151       Environment variables with the prefix PCP_ are used to parameterize the
152       file  and  directory names used by PCP.  On each installation, the file
153       /etc/pcp.conf contains the  local  values  for  these  variables.   The
154       $PCP_CONF  variable may be used to specify an alternative configuration
155       file, as described in pcp.conf(5).
156

SEE ALSO

158       PCPIntro(1),  pmdumplog(1),  pmlc(1),   pmlogextract(1),   pmlogger(1),
159       pcp.conf(5) and pcp.env(5).
160
161
162
163Performance Co-Pilot                  PCP                       PMLOGREDUCE(1)
Impressum