# $Id: README,v 1.2 2008/07/29 17:48:41 ksb Exp $

This program implements a shell level interface to the PEG
sampler host.  We take updates from a file (really stdin),
which we package then send to peg.

We expect updates in the form:
	update class/host/thing.rrd Ds1:Ds2:... Date:Sample2:Samlpe2:...\n
where:
	update		Is the literal word "update",
	class/host/thing.rrd
			Is the path from current working directory of
			the rrdd daemon we want to update,
	Ds1:Ds2:...	Are the names of the data sources (DS in RRD speak)
			we just sampled,
	Date		Can be either
		[Thu Nov 24 18:22:48 1986]
			A date in ctime(3) format w/o the "\n\000"
			enclosed in square brackets, or
		N	The string "N" for the present time
		1217353324
			An integral seconds since the epoch
	Sample2:Samlpe2:...
			The sample values that match the DS names sent
			in the order presented
	\n		is a newline character.

This is not the preferred way to send updates, as is forks another
process to do the (simple) UDP datagram formatting.  The better way to
to take the "sampler.pl" perl script as a template to code your own
sample that is persitant and takes all the std options.

If you can code a sampler as a function in perl (mkUpdate in the example
sampler) which returns the same text as rrdup(1) takes, without the word
"update " on the front, then you are in much better shape than starting
with a shell script and a pipe to rrdup.

--
ksb, July 2008