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

20030619: LDM 6.0.13 connections



Daniel,

>Date: Thu, 19 Jun 2003 17:36:19 -0400
>From: "Lemay,Daniel [CMC]" <address@hidden>
>Organization: Environment Canada
>To: "'Steve Emmerson'" <address@hidden>
>Subject: RE: 20030619: LDM 6.0.13 connections 

The above message contained the following:

> 1) It is normal that you are unable to connect with gilda.unidata.ucar.edu
> because the only address from unidata.ucar.edu  that can pass our firewall
> and CSS is shemp by virtue of the line:
> 
> allow ANY shemp.unidata.ucar.edu
> 
> in our ldmd.conf

Ah!

> 2) In attachment, I have include a log file in which you can see entries
> before we resolve a problem that has been caused by a misconfiguration of
> our CSS. Since the problem with the CSS is resolved, there is no more
> CLOSE_WAIT  but the connexion don't seem to be persistent because there is
> many TIME_WAIT that continue to appear.

The TIME_WAIT-s are due to the downsteam LDM-s connecting to your
top-level upstream LDM in order to ask it if the associated upstream LDM
is still alive.  This happens every 60 seconds when a downstream LDM
hasn't received anything from its upstream LDM.  If the downstream LDM
decides that its upstream LDM has died, then it closes the connection
and attempts to reconnect to the top-level, upstream LDM.

I'm trying to reduce the number of these disconnections and reconnections,
which is why I'd like to test my changes by connecting
to ldm.meteo.ec.gc.ca.  Is there any way you could add
".*\.unidata\.ucar\.edu" (or "gilda\.unidata\.ucar\.edu", at least) to
the set of allowed hosts?  I've configured the LDM on Gilda to not ask
for any data, so it should have a negligible effect on your system.

> [ldm@wxo-d1 logs]$ netstat -an | grep 388
> tcp        0      0 0.0.0.0:388             0.0.0.0:*               LISTEN
> tcp        0      0 10.10.10.112:388        128.117.140.27:43959
> ESTABLISHED
> tcp        0      0 10.10.10.112:388        128.118.28.12:47704
> ESTABLISHED
> tcp        0      0 10.10.10.112:388        128.118.28.12:48962
> ESTABLISHED
> tcp        0      0 10.10.10.112:388        128.117.13.119:57317
> TIME_WAIT
> tcp        0      0 10.10.10.112:388        128.117.140.27:53536
> TIME_WAIT
> tcp        0      0 10.10.10.112:388        128.117.13.119:41674
> ESTABLISHED
> tcp        0      0 10.10.10.112:388        132.208.133.165:50593
> TIME_WAIT
> tcp        0      0 10.10.10.112:388        128.118.28.12:57328
> TIME_WAIT
> tcp        0      0 10.10.10.112:388        132.208.133.165:44167
> ESTABLISHED

That's a lot nicer.

Regards,
Steve Emmerson