Re: [Testbed-admins] Event system questions

[Robert P Ricci <ricci@cs.utah.edu>]

Sorry, sent my last mail half-finshed. I can't answer all of your
questions, but I answer some....

Thus spake Robert P Ricci on Thu, Oct 01, 2009 at 04:48:18PM -0600:
> Thus spake Manuel Requena on Thu, Oct 01, 2009 at 01:40:05PM +0200:
> > When I swap in or modify an experiment, the events are not scheduled. I know
> > it because I launch some "tcpdump" in the nodes. Are there some log files to
> > see when the events are scheduled? I don't see my UDP packets (every 5ms
> > during 20 secs with a length of 500 bytes). But in the first 2 secs and
> > every 10ms, I see some UDP packets (not mine ones) of 62 bytes (ethernet
> > length) between the 2 nodes. What are these packets?
> 
> Hard to say - many switches send synchronization packets (for spanning
> tree, etc.), and this might be what they are - are the source MAC
> adresses ones the belong to the nodes themselves?
> 
> None of the Emulab software should be automatically sending packets on
> the experimental net (except for the traffic generators, if you have
> them set up...)
>  
> > If I "Replay Events" in the web interface, I can see my UDP packets. What I
> > have forgotten in order to launch the events when an experiment is created /
> > swapped in?

There is definitely something funny going on here, they should start on
swapin. I would check the following logs:

/proj/<pid>/exp/<eid>/logs/event-sched.log on ops, which has the event
scheulder log.

/var/log/emulab/ on the nodes themselves, which should have the logs for
the traffic generators themselves.
 
> > - In the "at" command, how the "0.0 time" is specified? When I "Replay
> > events", I guess that the "0.0 time" is more or less when I select the
> > option in the web interface. Could I specify the "0.0 time" at 13:15, for
> > example, and the rest of the "+XX" events are scheduled accordingly?

Time start occurs when all nodes finish booting, and all other times are
relative to that time. You *should* see a time start event in the event
scheduler log.

-- 
/-----------------------------------------------------------
| Robert P Ricci <ricci@cs.utah.edu> | <ricci@flux.utah.edu>
| Research Associate, University of Utah Flux Group
| www.flux.utah.edu | www.emulab.net
\-----------------------------------------------------------