On Sat, 4 Jul 2009, Tom Yoksas wrote:
Hi Gilbert, Ben, et. al.,
First, there were problems with the NEXRAD2 feed out of
idd.unidata.ucar.edu yesterday. That problem should have been
resolved yesterday evening.
That helped. I noticed that this morning. But then..
On Sat, 4 Jul 2009, Cotton, Benjamin J wrote:
So here's a followup to the previous message...
I chatted with Pat Finnegan, who runs pollux.itap.purdue.edu and
castor.itap.purdue.edu, Purdue's top-tier machines. Based on the rtstats
data, castor seems to be the only one with problems. Pat has switched
castor to feed from pollux instead of from the NWS. This seems to have
cleared up the latency problem. Pat mentioned he's been trying to get CR
to open up the firewall for a new machine (castor.rcac.purdue.edu), but
it has not been done yet. So my hypothesis is perhaps the CR firewall is
doing funny things to packets destined for castor.itap.purdue.edu. I'm
not sure what those funny things might be, but it at least fits the
situation.
So I guess the plan for now is to let pollux do all of the ingest from
the NWS and then both pollux and castor will serve the data to downstream
sites. I won't speak for Pat as to how long this arrangement might last,
but everyone should be getting their data on time now.
And that fixed a second problem I was having, and revealed to me a third
problem using pqmon: my queues weren't large enough to reject duplicates
delayed that long. So, I'm going to look at things in more detail on
Monday on how big my queues should be. Also, a data vendor I know of will
need to have their queues adjusted, more than likely.
Phew. Thanks, everyone, for your hard work. Now go enjoy some fireworks
and our freedom! :-)
*******************************************************************************
Gilbert Sebenste ********
(My opinions only!) ******
Staff Meteorologist, Northern Illinois University ****
E-mail: sebenste@xxxxxxxxxxxxxxxxxxxxx ***
web: http://weather.admin.niu.edu **
*******************************************************************************
