[Support #SCJ-652679]: ldm question

Michael,

> Good morning from KC.  I have an LDM question.  We're looking at bandwidth
> usage around here and since the data we supply to our field offices through
> LDM makes up a large part of the data we're passing, a need to measure that
> data has become necessary.  We can certainly measure data flow through our
> routers, but is there a way within LDM to look at the size of individual
> data files that make up the queue?  We want to measure at the individual
> file level so we can ask our field offices which pieces of data are
> necessary during an emergency when bandwidth demands would be high as
> opposed to regular business situations.

You have several options.  Most involve processing textual output:

1. The command "pqcat -vl- -i n >/dev/null" will print the metadata (including 
size) of data-products in the queue to the standard error stream, polling the 
queue every "n" seconds (without the "-i" option, it will run through the queue 
once and then terminate).

2. The command "ldmadmin watch" will do the same, polling every second.

3. The command "ldmadmin start -v" will cause the metadata of every received 
data-product to be written to the LDM log file.

4. Sending a USR2 signal to an upstream LDM process will cause it to log the 
metadata of every data-product to the LDM log file.

5. The pqact(1) entry "<feedtype><tab><pattern><tab>PIPE<tab>-nodata -metadata 
<program>" will pipe only the metadata of the data-product selected by 
<feedtype> and <pattern> to the program <program>.  This is an undocumented 
feature.  See the function "pipe_putmeta" in the file "pqact/filel.c" for 
details of the output.  I'm thinking about adding a "-text" option to have the 
output be textual rather than binary.  If you'd like this, then let me know.

Regards,
Steve Emmerson

Ticket Details
===================
Ticket ID: SCJ-652679
Department: Support LDM
Priority: Normal
Status: Closed