Alan.Hall wrote,
... With that in mind, are all of the servers in the network,
including the NWS Regional servers, configured with a large enough
queue to handle the increased load. If just one server somewhere in
the mix is configured too low, it could cause dropped data, and I have
seen this in summers past, but probably not duplicate data.
Wise words. At the Unidata Program Center, we take periodic snapshots
of relevant parameters (of which pqmon(1)'s age of the oldest product in
the queue is a very important one) and append them to a textual file
along with a timestamp. It's then a very simple matter to create
time-series plots of the parameters (using gnuplot(1) for example) to
determine if our system is configured correctly.
Most LDM sites should do something similar.
Regards,
Steve Emmerson
