SYNOPSISsheet2pcp [-h host] [-Z timezone] infile mapfile outfile
DESCRIPTIONsheet2pcp is intended to read a data spreadsheet (infile) translate this into a Performance Co-Pilot (PCP) archive with the basename outfile.
The input spreadsheet can be in any of the common formats, provided the appropriate Perl modules have been installed (see the CAVEATS section below). The spreadsheet must be normalized so that each row contains data for the same time interval, and one of the columns contains the date and time for the data in each row.
The resultant PCP archive may be used with all the PCP client tools to graph subsets of the data using pmchart(1), perform data reduction and reporting, filter with the PCP inference engine pmie(1), etc.
The mapfile controls the import process and defines the data mapping from the spreadsheet columns onto the PCP data model. The file is written in XML and conforms to the syntax defined in the MAPPING CONFIGURATION section below.
A series of physical files will be created with the prefix outfile. These are outfile.0 (the performance data), outfile.meta (the metadata that describes the performance data) and outfile.index (a temporal index to improve efficiency of replay operations for the archive). If any of these files exists already, then sheet2pcp will not overwrite them and will exit with an error message.
The -h option is an alternate to the hostname attribute of the <sheet> element in mapfile described below. If both are specified, the value from mapfile is used.
The -Z option is an alternate to the timezone attribute of the <sheet> element in mapfile described below. If both are specified, the value from mapfile is used.
sheet2pcp is a Perl script that uses the PCP::LogImport Perl wrapper around the PCP libpcp_import library, and as such could be used as an example to develop new tools to import other types of performance data and create PCP archives.
MAPPING CONFIGURATIONThe mapfile contains specifications in standard XML format.
The whole specification is wrapped in a <sheet> ... </sheet> element. The sheet tag supports the following optional attributes:
- Specifies the number of heading rows to skip at the start of the spreadsheet before processing data. Example: heading=1.
- Set the source hostname in the PCP archive (the default is to use the hostname of the local host). Example: hostname=some.where.com.
- Set the source timezone in the PCP archive (the default is to use UTC). The timezone must have the format +HHMM (for hours and minutes East of UTC) or -HHMM (for hours and minutes West of UTC). Note in particular that neither the zoneinfo (aka Olson) format, e.g. Europe/Paris, nor the Posix TZ format, e.g. EST+5 is allowed. Example: timezone=+1100.
- The format of the date imported from the spreadsheet may be specified as a concatenation of values that specify the order of the year (Y), month (M) and day (D) fields in a date. The supported variants are DMY (the default), MDY and YMD. Example: datefmt=YMD.
A <sheet> element contains one or more metric specifications of the form <metric>metricname</metric>. The metric tag supports the following optional attributes:
- The Performance Metrics Identifier (PMID), specified as 3 numbers separated by a periods (.) to set the domain, cluster and item fields of the PMID, see PMNS(5) for more details of PMIDs. If omitted, the PMID will be automatically assigned by pmiAddMetric(3). The value PM_ID_NULL may be used to explicitly nominate the default behaviour. Examples: pmid=60.0.2, pmid=PM_ID_NULL.
- Each metric may have one or more values. If a metric always has one value, it is singular and the Instance Domain should be set to PM_INDOM_NULL. Otherwise indom should be specified as 2 numbers separated by a period (.) to set the domain and ordinal fields of the Instance Domain, see the __pmInDom_int typedef in <pcp/impl.h>. Examples: indom=PM_INDOM_NULL, indom=60.3, indom=PMI_DOMAIN.4. More than one metric can share the same Instance Domain when the metrics have defined values over similar sets of instances, e.g. all the metrics for each network interface. It is standard practice for the domain field to be the same for the pmid and the indom; if the pmid attribute is missing, then the domain field for the indom should be the reserved domain PMI_DOMAIN. If the indom attribute is omitted then the default Instance Domain for the metric is PM_INDOM_NULL.
- The scale and dimension of the metric values along the axes of space, time and count (events, messages, packets, etc.) is specified with a 6-tuple. These values are passed to the pmiUnits(3) function to generate a pmUnits structure. Refer to pmLookupDesc(3) for a full description of all the fields of this structure. The default is to assign no scale or dimension to the metric, i.e. units=0,0,0,0,0,0. Examples: units=0,1,0,0,PM_TIME_MSEC,0 (milliseconds), units=1,-1,0,PM_SPACE_MBYTE,PM_TIME_SEC,0 (Mbytes/sec), units=0,1,-1,0,PM_TIME_USEC,PM_COUNT_ONE (microseconds/event).
- Defines the data type for the metric. Refer to pmLookupDesc(3) for a full description of the possible type values; the default is PM_TYPE_FLOAT. Examples: type=PM_TYPE_32, type=PM_TYPE_U64, type=PM_TYPE_STRING.
- Defines the semantics of the metric. Refer to pmLookupDesc(3) for a full description of the possible values; the default is PM_SEM_INSTANT. Examples: sem=PM_SEM_COUNTER, type=PM_SEM_DISCRETE.
The remaining specifications define the data columns in order using exactly one <datetime></datetime> element, one or more <data>metricspec</data> elements and one or more <skip></skip> elements.
The <datetime> element defines the column in which a date and time will be found to form the timestamp in the PCP archive for all the data in each row of the PCP archive.
For the <data> element, a metricspec consists of a metric name (as defined in an earlier <metric> element), optionally followed by an instance name that is enclosed by square brackets, e.g. <data>hinv.ncpu</data>, <data>kernel.all.load[1 minute]</data>.
The skip tag defines the column that should be skipped when preparing data for the PCP archive.
The order of the <datetime>, <data> and <skip> elements matches the order of columns in the spreadsheet. If the number of elements is not the same as the number of columns a warning is issued, and the extra elements or columns generate no metric values in the output archive.
EXAMPLEThe mapfile ...
<?xml version="1.0" encoding="UTF-8"?> <sheet heading="1"> <!-- simple example --> <metric pmid="60.0.2" indom="60.0" units="0,1,0,0,PM_TIME_MSEC,0" type="PM_TYPE_U64" sem="PM_SEM_COUNTER"> kernel.percpu.cpu.sys</metric> <datetime></datetime> <skip></skip> <data>kernel.percpu.cpu.sys[cpu0]</data> <data>kernel.percpu.cpu.sys[cpu1]</data> </sheet>
could be used for a spreadsheet in which the first few rows are ...
Date;"Status";"SysTime - 0";"SysTime - 1"; 26/01/2001 14:05:22;"Some Busy";0.750;0.133 26/01/2001 14:05:37;"OK";0.150;0.273 26/01/2001 14:05:52;"All Busy";0.733;0.653
CAVEATSOnly the first sheet from infile will be processed.
Additional Perl modules must be installed for the various spreadsheet formats, although these are checked for ar run-time so only the modules required for the specific types of spreadsheets you wish to process need be installed:
- Spreadsheets in the Comma Separated Values (CSV) format require Text::CSV_XS(3pm).
- *.sxc or *.ods
- OpenOffice documents require Spreadsheet::ReadSXC(3pm), which in turn requires Archive::Zip(3pm).
- Classical Microsoft Office documents require Spreadsheet::ParseExcel(3pm), which in turn requires OLE::Storage_Lite(3pm).
- Microsoft OpenXML documents require Spreadsheet::XLSX(3pm). sheet2pcp does not appear to work with OpenXML documents saved from OpenOffice.