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

20050519: LDM sequencing

Michael,

>Date: Thu, 19 May 2005 10:18:49 -0500
>From: Michael McEniry <address@hidden>
>Organization: University of Utah
>To: Steve Emmerson <address@hidden>
>Subject: Re: 20050519: LDM sequencing

The above message contained the following:

> We seem to get a lot more connections (hence, a lot more 
> forks) than I would have expected. From the log entries, 
> these seem to be ISALIVE queries.
> 
> Based on the behavior docs, it seems that each FEEDME 
> connection sticks around until a corresponding ISALIVE query 
> fails, and that each ISALIVE query is a separate connection. 
> Is that correct?

That's correct.  A downstream LDM will establish a connection to the
top-level, upstream, LDM server for an ISALIVE inquiry if the downstream
LDM hasn't received anything in one minute.  To prevent this, an
upstream LDM sends an empty packet to the downstream LDM every 30
seconds.

> Most of the data running through this system 
> (ldm.itsc.uah.edu) is forecast model related, so it arrives 
> in big groups every few hours. Most of the time, there's no 
> data to send. At least one of our downstream clients reports 
> usually missing some products from each group. I'm trying to 
> sort through the ldmd.log to figure out which entries are 
> important and which aren't (ISALIVE and "End of product-queue").

Is the downstream LDM reconnecting?  It is possible for a downstream LDM
to miss data-products this way -- especially if the system clocks on the
upstream LDM and whatever host created the data-products aren't
synchronized.  This will be fixed in the next release.

The "Format and Interpretation" page of the "LDM Basics" webpages might
be useful:

    
http://my.unidata.ucar.edu/content/software/ldm/ldm-6.3.0/basics/logfile-format.html

Regards,
Steve Emmerson
LDM Developer