[hobbit] All hosts network traffic graph page

[Henrik Stoerner]

On Sat, Mar 08, 2008 at 09:35:49PM -0600, Taylor Lewick wrote:
> Is there a simple way to setup a view via a report or link in hobbit so
> that I can display all of the hosts network traffic graphs?
> 
> Kind of like an mrtg style view.  That way if we see bandwidth spike on
> our switch we can pull up this page and get a quick view of which hosts
> are talking a lot?

Tricky, because network interface names vary from one host to another.

What you can do is to create a multi-graph definition in 
hobbitgraph.cfg that picks out the exact NIC graph you want. 
Copy the existing "ifstat" definition, and make a new "eth0-multi"
graph where the FNPATTERN refers only to the interface graph 
you want - i.e. instead of the "ifstat.(.+).rrd" pattern you 
refer to the exact "ifstat.INTERFACE.rrd" file matching the name
of your NIC. E.g. for "eth0":

[eth0-multi]
        FNPATTERN ifstat.eth0.rrd
        TITLE Network Traffic
        YAXIS Bits/second
        DEF:inbytes at RRDIDX@=@RRDFN@:bytesReceived:AVERAGE
        CDEF:in at RRDIDX@=inbytes at RRDIDX@,8,*
        DEF:outbytes at RRDIDX@=@RRDFN@:bytesSent:AVERAGE
        CDEF:out at RRDIDX@=outbytes at RRDIDX@,8,*
        LINE2:in at RRDIDX@#@COLOR@:@RRDPARAM@  inbound
        GPRINT:in at RRDIDX@:LAST: \: %10.1lf (cur)
        GPRINT:in at RRDIDX@:MAX: \: %10.1lf (max)
        GPRINT:in at RRDIDX@:MIN: \: %10.1lf (min)
        GPRINT:in at RRDIDX@:AVERAGE: \: %10.1lf (avg)\n
        LINE2:out at RRDIDX@#@COLOR@:@RRDPARAM@ outbound
        GPRINT:out at RRDIDX@:LAST: \: %10.1lf (cur)
        GPRINT:out at RRDIDX@:MAX: \: %10.1lf (max)
        GPRINT:out at RRDIDX@:MIN: \: %10.1lf (min)
        GPRINT:out at RRDIDX@:AVERAGE: \: %10.1lf (avg)\n

Then you can create a new metric on the "Metrics Report" page by adding
a new "<option>" ... in the ~hobbit/server/web/hostgraphs_form : Find 
the "<SELECT" tag and add an option for the "eth0" graph:

	<SELECT NAME=testname MULTIPLE SIZE=15>
	<option value="ALL" SELECTED>ALL</option>
	... other current options ...
	<option value="eth0">Network I/O eth0</option>

Note that in hobbitgraph.cfg the name is "eth0-multi", but in the
hostgraphs_form it is just "eth0". The "-multi" is added automatically.

You should then be able to view the eth0 I/O from the Metrics report
page.


Regards,
Henrik