[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[LDM #AET-785898]: primary/alternate

Karen,

> request EXP     ".*" 172.18.34.30
> request EXP     ".*" 172.18.34.31
> request EXP     ".*" 172.18.34.32
> request EXP     ".*" 172.18.34.34
> request WMO     ".*" 172.18.34.30
> request WMO     ".*" 172.18.34.31
> request WMO     ".*" 172.18.34.32
> request WMO     ".*" 172.18.34.34
> request NMC2    ".*" 172.18.34.30
> request NMC2    ".*" 172.18.34.31
> request NMC2    ".*" 172.18.34.32
> request NMC2    ".*" 172.18.34.34
> request DDS     ".*" 172.18.34.29
> request WMO     ".*" 172.18.34.29
> request EXP     ".*" 172.18.34.35
> request WMO     ".*" 172.18.34.35
> request NMC2    ".*" 172.18.34.35
> request WMO     ".*" 172.18.34.87

The REQUEST entries might cause unnecessary disconnections.  In judging
its insertion success rate, a downstream LDM knows how many other
downstream LDM-s are receiving the same stream of product.  The
feedtype and pattern are used to determine the number of such
processes.  In your REQUEST entries, you have multiple requests for the
same feedtype and pattern.  Unfortunately, the received streams are all
unique, which violates the assumption of the switching algorithm.

I suggest using the following entries:

request EXP        .* 172.18.34.30
request EXP        (.*) 172.18.34.31
request EXP        ((.*)) 172.18.34.32
request EXP        (((.*))) 172.18.34.34
request EXP        ((((.*)))) 172.18.34.35
request WMO      .* 172.18.34.30
request WMO      (.*) 172.18.34.31
request WMO      ((.*)) 172.18.34.32
request WMO      (((.*))) 172.18.34.34
request WMO      ((((.*)))) 172.18.34.29
request WMO      (((((.*))))) 172.18.34.35
request WMO      ((((((.*)))))) 172.18.34.87
request NMC2     .* 172.18.34.30
request NMC2     (.*) 172.18.34.31
request NMC2     ((.*)) 172.18.34.32
request NMC2     (((.*))) 172.18.34.34
request NMC2     ((((.*)))) 172.18.34.35
request DDS        .* 172.18.34.29

Adding different number of matching parentheses around the pattern
ensures that each data stream is considered unique.

If you can modify the patterns so that they match the data-products
in each stream but are really unique (e.g., "^A", "^B", etc.) then that
would be even better because a downstream LDM searches back through
the product-queue after a disconnect to find the most recently
received data-product that matches its product-class -- and adding
parentheses won't change that.

This is kind of esoteric and advanced, so feel free to ask questions.

Regards,
Steve Emmerson

Ticket Details
===================
Ticket ID: AET-785898
Department: Support LDM
Priority: Normal
Status: Closed