[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: 20040806: GEMPAK/LDM problem...
- Subject: Re: 20040806: GEMPAK/LDM problem...
- Date: 10 Aug 2004 16:12:07 -0600
Robert,
You would restrict the data you request in your ldmd.conf file,
so that instead of a request line specifying CONDUIT ".*", you
would create a pattern of the data you are interested in.
For example, if you wanted the ETA grid 212, the AVN/GFS through 84
hours, and all RUC, you would use a pattern such as:
"MT.eta.*#212|MT.avn.*fh.00([0-7].|8[0-4])|MT.ruc"
Steve Chiswell
Unidata User SUpport
On Sun, 2004-08-08 at 18:50, Robert Dewey wrote:
> Hello,
>
> I figured out part of the problem, and it was related to connection
> problems. I have 2 network cards on the same machine, and both were
> using the same MAC address... This caused quite a bit of packet loss,
> with connection speeds dropping as low as 56K!
>
> I am still having CONDUIT latency greater than 6000 seconds (I believe
> data is stored up to 3600 seconds, so anything past that would be long
> gone), but I seem to have all of the data without any losses. At the
> moment, I am only requesting a few model grids (my hourly average is
> ~328MB for CONDUIT) . I believe it is the AVN/GFS which may be causing
> the latency though, since these files get as large as 400-500MB... Is
> there something I could add into my pqact.conf to only allow data out to
> 84 hours? I am kind of new to the LDM, and not exactly sure how/if this
> could be done...
>
> Thanks!
> Robert Dewey
>
> On Sun, 2004-08-08 at 17:36, Steve Chiswell wrote:
> > The most likely problems would be you don't have enough bandwidth to try to
> > receive all of Conduit (which can be as much as 1.8 GB an hour).
> >
> > Looking at the reception stats for what I believe is your LDM:
> >
> > http://my.unidata.ucar.edu/cgi-bin/rtstats/iddstats_nc?CONDUIT+d14-69-10-190.try.wideopenwest.com
> >
> > you see frequent excursions even on the weekend where your latencies are so
> > large that your upstream likely no longer has the data by the time your LDM
> > could receive the data. You have values as large as 10,000 seconds, and even
> > today will see peaks during the main model times around 4000 seconds
> > latency.
> >
> > The best bet is to reduce the amount of data you are requesting to fit
> > within
> > your limitations.
> >
> > Steve Chiswell
> > Unidata User Support
> >
> >
> > On Fri, 6 Aug 2004, Robert Dewey wrote:
> >
> > > Hello,
> > >
> > > I am having some troubles with the LDM, particularly the CONDUIT feed. I
> > > am always missing random grids from a gridded set (ETA/GFS/etc.). Here
> > > is an example of the 06Z ETA from this morning, I used GARP to loop
> > > hours 00 through 84:
> > >
> > > GEMPAK: [DG -7] Input grid EMSL ^040806/0600F063 @0 %NONE cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid EMSL ^040806/0600F069 @0 %NONE cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid EMSL ^040806/0600F084 @0 %NONE cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid EMSL ^040806/0600F063 @0 %NONE cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid EMSL ^040806/0600F069 @0 %NONE cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid EMSL ^040806/0600F084 @0 %NONE cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid WND ^040806/0600F057 @500 %PRES cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid WND ^040806/0600F060 @500 %PRES cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid WND ^040806/0600F063 @500 %PRES cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid WND ^040806/0600F069 @500 %PRES cannot be
> > > found.
> > > GEMPAK: [DG -7] Input grid WND ^040806/0600F084 @500 %PRES cannot be
> > > found.
> > >
> > > As you can see, I am missing random grids. What could the problem be? I
> > > have set my product queue in the ldmadmin script to 2GB (thinking that
> > > maybe the data is being lost), with my physical memory being 1GB (I have
> > > an additional 2GB of SWAP, but this never gets used, the entire system
> > > usually never uses more than 900MB). I do not have any "accept" lines in
> > > my ldmd.con file, but I don't think that is the problem. I checked my
> > > ldmd.log files, and found these mixed in (quite frequently):
> > > _________________________________________________________________________________________________
> > > Aug 06 22:03:38 LDM weather2[2648]: Connected to upstream LDM-6
> > > Aug 06 22:03:38 LDM weather2[2648]: Upstream LDM is willing to feed
> > > Aug 06 22:11:47 LDM bigbird[2725]: ERROR: requester6.c:206: Connection
> > > to upstream LDM closed
> > > Aug 06 22:12:02 LDM weather2[2649]: ERROR: requester6.c:206: Connection
> > > to upstream LDM closed
> > > Aug 06 22:12:17 LDM bigbird[2725]: Desired product class:
> > > 20040806221107.220 TS_ENDT {{CRAFT, ".(KDTX|KGRR|KIWX)"}}
> > >
> > > Aug 06 22:47:39 LDM weather2[2646]: Connected to upstream LDM-6
> > > Aug 06 22:47:39 LDM weather2[2646]: Upstream LDM is willing to feed
> > > Aug 06 22:48:46 LDM bigbird[2650]: ERROR: requester6.c:206: Connection
> > > to upstream LDM closed
> > > Aug 06 22:49:16 LDM bigbird[2650]: Desired product class:
> > > 20040806213319.386 TS_ENDT {{CONDUIT, ".*"}}
> > > _________________________________________________________________________________________________
> > >
> > > I checked the RUC2 data, from the CONDUIT feed... And all of the grids
> > > seem to be there, so I am not sure what exactly the problem is, or what
> > > these error messages mean...
> > >
> > > Here is the LDM system stats:
> > >
> > > O/S: Fedora Core 2
> > > Processor: 1.8Ghz AMD Duron
> > > System RAM: 1GB DDR-2100 RAM, (2GB SWAP)
> > > Hard Drive: 80GB HDD (7200RPM)
> > > Connection: 3-4Mbps connection
> > >
> > > Thanks!
> > > Robert Dewey
> > >
> > >
> > >
> >