pmlogreduce - temporal reduction of Performance Co-Pilot archives
$PCP_BINADM_DIR/pmlogreduce [-z] [-A align] [-S starttime] [-s samples] [-T endtime] [-t
interval] [-v volsamples] [-Z timezone] input output
pmlogreduce reads one Performance Co-Pilot (PCP) archive identified by input (this must be
a PCP archive created by pmlogger(1), pmlogextract(1) or pmlogreduce(1)), and creates a
temporally reduced PCP archive in output. The data reduction involves statistical and
temporal reduction of samples with an output sampling interval defined by the -t option in
the output archive (independent of the sampling intervals in the input archive), and is
further controlled by other command line arguments.
For some metrics, temporal data reduction is not going to be helpful, so for metrics with
types PM_TYPE_AGGREGATE or PM_TYPE_EVENT, a warning is issued if these metrics are found
in input and they will be skipped and not appear in the output archive.
COMMAND LINE OPTIONS
The command line options for pmlogreduce are as follows:
Specify a ``natural'' alignment of the output sample times; refer to PCPIntro(1).
Define the start of a time window to restrict the samples retrieved from the input
archive; refer to PCPIntro(1).
The argument samples defines the number of samples to be written to output. If
samples is 0 or -s is not specified, pmlogreduce will sample until the end of the
PCP archive, or the end of the time window as specified by -T, whichever comes
first. The -s option will override the -T option if it occurs sooner.
Define the termination of a time window to restrict the samples retrieved from the
input archive; refer to PCPIntro(1).
The output archive is potentially a multi-volume data set, and the -v option causes
pmlogreduce to start a new volume after volsamples log records have been written to
the output archive.
Independent of any -v option, each volume of an archive is limited to no more than
2^31 bytes, so pmlogreduce will automatically create a new volume for the archive
before this limit is reached.
Consecutive samples in the output archive will appear with a time delta defined by
interval; refer to PCPIntro(1). Note the default value is 600 (seconds, i.e. 10
Use timezone when displaying the date and time, or interpreting the -S and -T
options. Timezone is in the format of the environment variable TZ as described in
-z Use the local timezone of the host from the input archive when displaying the date
and time, or interpreting the -S and -T options. The default is to initially use
the timezone of the local host.
The statistical and temporal reduction follows the following rules:
1. Consecutive records from input are read without interpolation, and at most one output
record is written for each interval, summarizing the performance data over that
2. If the semantics of a metric indicates it is instantaneous or discrete then output
value is computed as the arithmetic mean of the observations (if any) over each inter-
3. If the semantics of a metric indicates it is a counter then the following transforma-
tions are applied:
a) Metrics with 32-bit precision are promoted to 64-bit precision.
b) Any counter wrap (overflow) is noted, and appropriate adjustment made in the value
of the metric over each interval. This will be correct in the case of a single
counter wrap, but will silently underestimate in the case where more than one
counter wrap occurs between consecutive observations in the input archive, and
silently overestimate in the case where a counter is reset occurs between consecu-
tive observations in the input archive; unfortunately these situations cannot be
detected, but are believed to be rare events for the sort of production monitoring
environments where pmlogreduce is most likely to be deployed.
4. Any changes in instance domains, and indeed all metadata, is preserved.
5. Any ``mark'' records in the input archive (as created by pmlogextract(1)) will be pre-
served in the output archive, so periods where no data is available are maintained,
and data interpolation will not occur across these periods when the output archive is
subsequently processed with PCP applications.
For each of the input and output archives, several physical files are used.
metadata (metric descriptions, instance domains, etc.) for the archive log
archive.0 initial volume of metrics values (subsequent volumes have suffixes 1, 2, ...) -
for input these files may have been previously compressed with bzip2(1) or
gzip(1) and thus may have an additional .bz2 or .gz suffix.
temporal index to support rapid random access to the other files in the archive
Environment variables with the prefix PCP_ are used to parameterize the file and directory
names used by PCP. On each installation, the file /etc/pcp.conf contains the local values
for these variables. The $PCP_CONF variable may be used to specify an alternative config-
uration file, as described in pcp.conf(5).
PCPIntro(1), pmdumplog(1), pmlc(1), pmlogextract(1), pmlogger(1), pcp.conf(5) and
All error conditions detected by pmlogreduce are reported on stderr with textual (if some-
times terse) explanation.
Should the input archive be corrupted (this can happen if the pmlogger instance writing
the archive suddenly dies), then pmlogreduce will detect and report the position of the
corruption in the file, and any subsequent information from the input archive will not be
If any error is detected, pmlogreduce will exit with a non-zero status.
The preamble metrics (pmcd.pmlogger.archive, pmcd.pmlogger.host, and pmcd.pmlogger.port),
which are automatically recorded by pmlogger at the start of the archive, may not be
present in the archive output by pmlogreduce. These metrics are only relevant while the
archive is being created, and have no significance once recording has finished.
Performance Co-Pilot PCP PMLOGREDUCE(1)