[AG-TECH] Steal the Beacon
Marty.Hoag at ndsu.nodak.edu
Thu Apr 25 17:28:48 CDT 2002
Ok, who's killing off the beacon clients?
I noticed earlier today that a couple favorites at ANL were
not on the beaconserver (lib-video and ws-video) while two others
(aem, aglintest). Bob Olson said they were running but were not
sending any of the unicast reports to the beaconserver.
Then this afternoon I noticed my own beacon had disappeared from
the subset I watch regularly! Figuring I couldn't harass others
without my beacon reporting I checked with Bob and again, he say
no reports coming to the beaconserver. HOWEVER, I check the Abilene
IPLS router and I was seeing the normal flow of 9 packets per second
on the state entry for our beacon (184.108.40.206,
We apparently didn't have tcpdump on the audio machine under the
AGiB build so I rebooted the machine to make sure I got a complete
shutdown and restart (gun shy after seeing too many nodes with
multiple beacons running). That did the trick - Bob immediately
saw reports coming back in (this was around 1700 CDT/2200 UTC) and
in a couple minutes we again appeared on the beacon pages.
So, is our warranty up on the v0.63 beacon clients? Did someone
forget to put a nickle in the java runtime slot? Is someone (Mitch,
you around) testing a new beacon and killing off the reporting process?
We're down to 112 beacons after having been over 120 for some days
so maybe this has happened to others. Or maybe this is an old problem
now that I think about it as in the past I've harassed folks about
their beacon running and they swore they had one going. And maybe
that is how duplicates get started - no reports inplies no beacon
running so we just start another one which then creates twice the
multicast traffic and those intermittent 99 entries.
It does point up that it might be nice to get some reports from
the server on things like "beacons heard by other beacons but not
reporting" (or maybe showing the beacon but somehow indicating in
the beacon info that no report was heard from the beacon directly,
all information is from the other clients), beacons being rejected
for "bad" content, the last date/time/time zone when each beacon
was "first heard" as far as the unicast udp reports to the server,
I haven't seen much of Mitch lately (maybe it was something I
said?) but it would be interesting to figure out if several sites
losing beacon reports around the same times is coincidence or what.
More information about the ag-tech