The important thing to remember is that the maximum latency parameter
in the registry should be consonant with the minimum residence time of
the queue in order to support duplicate product detection and
rejection.
Be advised that when the reconciliation-mode registry parameter isn't
"do nothing", the command "ldmadmin vetqueuesize" is kinda like
cruise-control in a car: you don't want to engage it until you're close
to where you want to be.
We run 180 GB queues here, but we get everything and feed many.
Hiya Steve.. good to see your name again :)
So if you run 180GB queues at ucar, you clearly have a reason for doing
so.. could you expand on that logic please? :)
Surely supporting duplicate product detection would be one reason, and
perhaps the ability to support downstream feeds if they need to rebuild
their own queues during periods of unforseen downtime?
My largest queue to date has been 32GB with the coal of conservatism,
and the ability to quickly reseed my downstream production servers /
feeds :)
cheers,
--patrick
