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

Re: 19991103: LDM 5.0.8 and 'expanding input buffer size' problemwithAFOS feed

Gregory,

I don't have any good solutions to your problem.  I created an source code
release for ldm-5.0.2 in the standard ldm source ftp directory.  You could
compile the source, test it and then modify the PIL fixes to the code. One
word of caution, save all the other working releases so no overwrites
occur.  At this time, I can't really expend any more time because AFOS
stream is not really used in the IDD. If I think of any new ideas about
the problems, I'll let you know.  I had numerous problems with HPUX, if
fact I can't get it compiled on HPUX 11.0 because of a bug in the rpc
library. For HPUX 11.0 users, they have to use the version compiled on
HPUX 10.2   I sympathize with your situation, it's frustrating.


Robb...




On Thu, 4 Nov 1999, Gregory Grosshans wrote:

> I'm running LDM 5.0.2, and trying to get 5.0.8 to work on an HP 712 HPUX 10.20
> machine.  Up until this past spring the LDM was running on an HP 715 with 
> HPUX 9.x.
> The feeds into this machine are AFOS, DDS and PPS via serial connections.
> 
> I just grabbed the HPUX 10.2 LDM 5.0.8 binaries off the web site and 
> installed them and
> set the queue at 100 MB.  Again I received
> the "Expanding input buffer size" problem.
> 
> I've attached ldmd.log.1 and ldmd.log.2.  ldmd.log.1 was created when I tried 
> to run
> the 5.0.8 binaries and you will notice the
> expanding buffer size messages.  The interesting item at the end of this log 
> is under
> afos it lists 'nregions    35'.  While the PPS and
> DDS feeds don't list an nregions at all.  Then if you look at the end of 
> ldmd.log.2
> (created from 5.0.2) under AFOS it doesn't list
> 'nregions...'.    Do you know what causes LDM to generate this message?
> 
> Outside of just a few HP's with Mux ports the other alternatives are 
> configuring a
> linux box.   Operations is just an HP shop.
> 
> The router configuration would involve NCEP HQ so I'd like to wait until 
> absolutely
> necessary to go that route.
> 
> Gregg
> 
> 
> 
> Robb Kambic wrote:
> 
> > Gregory,
> >
> > At this point, I'm really starting to speculate. Is it possible to install
> > the ldm on another box that's not IRIX?  My thinking is some library/etc
> > has changed to make the ldm behave abnormal. If we could duplicate this on
> > another platform then it's definitely a bug in the ldm.  Could be OS
> > specific.
> >
> > Robb...
> >
> > On Thu, 4 Nov 1999, Gregory Grosshans wrote:
> >
> > > On the machine in question the only feeds are DDS, PPS and AFOS.  On this 
> > > machine
> > > I've had a LDM queue of 25 MB for three years and it worked fine with all 
> > > three
> > > feed types.
> > >
> > > There is a seperate machine with LDM ingesting NOAAPORT and on this 
> > > machine the
> > > queue is 700+ MB since I'm ingesting three
> > > channels of NOAAPORT data.
> > >
> > > I've actually set the size of the ldm queue in the $LDMHOME/bin/ldmadmin 
> > > script to
> > > 250000000.  I've verified several times that the queue is of sufficient 
> > > length.
> > > When I switch back to 5.0.2 I change the queue back to 25 MB and 
> > > everything works
> > > fine.
> > >
> > > I tried commenting out the PPS and DDS feed types, leaving only AFOS on 
> > > and
> > > starting 5.0.8 and still received the 'expanding input buffer size'.  
> > > While LDM
> > > was running with just the AFOS feed I tried pqcat.  PQCAT started and 
> > > exited right
> > > away and indicated the number of products were 0.
> > >
> > > Gilbert, If I were able to have the routers and firewall changed to allow 
> > > your LDM
> > > access to this machine I don't understand how the AFOS
> > > data would be able to get to you.  Right now there are no AFOS products 
> > > entering
> > > the queue, the only thing occurring is the 'Expanding input buffer size.' 
> > >  If
> > > there were AFOS products entering the queue they would also show up in 
> > > one of the
> > > three other LDM machines here.
> > >
> > > Actually, that may be a key, the buffer is expanding.  Robb, do you know 
> > > how the
> > > buffer plays into the LDM?   Does LDM load a product
> > > into a new dynamic buffer prior to placing it in the queue and then does 
> > > it free
> > > the memory?\
> > >
> > > Thanks for your timely responses.
> > >
> > > Gregg
> > >
> > >
> > > Gilbert Sebenste wrote:
> > >
> > > > On Thu, 4 Nov 1999, Robb Kambic wrote:
> > > >
> > > > > Gregory,
> > > > >
> > > > > After looking at the log messages, it doesn't appear your queue is 250
> > > > > megabytes.  Look in the data dir, what's the:
> > > > >
> > > > > % ls -l ldm.pq
> > > > >
> > > > > If it's 250 megabytes then up it again. There might be other products
> > > > > coming in on your afos stream, etc.  If it isn't then change it in the
> > > > > bin/ldmadmin file to 250.  Then the standard
> > > > >
> > > > > % ldmadmin stop
> > > > > % ldmadmin delqueue
> > > > > % ldmadmin mkqueue
> > > > > % ldmadmin start
> > > > >
> > > > > I just put the commands here for my sanity sake.
> > > >
> > > > DUH! I forgot he was also ingesting NOAAPORT. Whoa, that's 175 MB right
> > > > there. Up, up and away! Use 400 MB at least.
> > > >
> > > > > Also, if you comment out the dds and pps does it make any difference?
> > > > >
> > > > > Robb...
> > > >
> > > > I think that's the key here. NOAAPORT takes up 150 MB of my queue/hour,
> > > > and of course, we filter out scrambled NIDS and other stuff. Give it a 
> > > > go,
> > > > and then let us know!
> > > >
> > > > *******************************************************************************
> > > > Gilbert Sebenste                                                     
> > > > ********
> > > > Internet: address@hidden    (My opinions only!)                     
> > > > ******
> > > > Staff Meteorologist, Northern Illinois University                      
> > > > ****
> > > > Work phone: 815-753-5492                                                
> > > > ***
> > > > *******************************************************************************
> > > >
> > >
> >
> > ===============================================================================
> > Robb Kambic                                Unidata Program Center
> > Software Engineer III                      Univ. Corp for Atmospheric 
> > Research
> > address@hidden                   WWW: http://www.unidata.ucar.edu/
> > ===============================================================================
> 

===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================