20041028: Solon Latencies

>From:  David Garrana Coelho <address@hidden>
>Organization:  UFRJ
>Keywords:  200410290154.i9T1skvV005422 IDD ldm-6.1.0

Hi David,

>       Thank you very much for the tips. Combine did the trick, at least 
>for the multichannel images. Not sure how to make the topo/IR/VIS ones 
>yet.

The procedure for doing the combined topo/IR/VIS images is included in
the McIDAS distribution and ready to run on any machine running the LDM
that is getting the UNIWISC images and has McIDAS installed.  The
generation of these images is active on solon, so you can see the
process "live".  We can talk about what would be needed to get the same
thing running for South American images when I get some moments.

>       Since last week we are having very bad latency problems with 
>solon/LDM, which in turn is damaging the GFS files we get from CONDUIT. 
>Some of them are even bigger than the usual, which i fixed by removing 
>Thelma from the alternate sources. Yet, the files are getting somehow 
>damaged (smaller) than expected. This is halting our model runs, and 
>product generation since last saturday. I know you must be uber busy (as 
>usual), but there is a way to check if its a local (UFRJ) or external 
>network problem? I need to be sure the problem is local before contact 
>network support, because I know that the first thing they will argue is 
>that the Int'l link to blame.
>       I suspected that LDM 6.1.0 could be the villain here, but it's 
>very unlikely, right?

Steve Emmerson commented to me yesterday morning that he found and
fixed a subtle problem in the LDM that I thought was just related to
AIX systems The problem manifested itself in a decrease in ingested
data when rpc.ldmd connections were unexpectedly terminated.  If the
connection between solon and its upstream feed host(s) is dropping, the
problem that Steve found may go a long way to helping you out.  I must
say quickly that I do _not_ know that this is the problem you are
seeing on solon, but it may be.  I will ask Steve this morning if
he can cut a test version of the new code that we can try out on
solon.

In the meantime, since ldm-6.0.14 had been working very well for a long
time, I decided to switch solon back to using it -- just to make sure
we are not seeing a problem with LDM-6.1.0.

>       BTW, you can check file sizes at:
>       http://solon.meteoro.ufrj.br/dados/conduit

I will contact you about a the new test version of the LDM after I have
had a chance to talk to Steve.

>       Thanks in advance (again),

No worries.

Cheers,

Tom
--
+-----------------------------------------------------------------------------+
* Tom Yoksas                                             UCAR Unidata Program *
* (303) 497-8642 (last resort)                                  P.O. Box 3000 *
* address@hidden                                   Boulder, CO 80307 *
* Unidata WWW Service                             http://www.unidata.ucar.edu/*
+-----------------------------------------------------------------------------+