Re: 19990727: ldmadmin isrunning (fwd)

Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
rkambic@xxxxxxxxxxxxxxxx                   WWW:

---------- Forwarded message ----------
Date: Thu, 29 Jul 1999 11:46:00 -0600
From: Jim Cowie <cowie@xxxxxxxxxxxx>
To: Robb Kambic <rkambic@xxxxxxxxxxxxxxxx>
Subject: Re: 19990727: ldmadmin isrunning

Robb Kambic wrote:


Another solution would be to add the command ldmadmin clean to the start
up script, that clears out the lock file and the file.  I decided
to go with your solution because it doesn't cause any changes to the LDM
preinstall configurations.

Yes, the ldmadmin clean would work if this is used in a boot-up
script, but I actually do this from a cron job that runs every half
hour to check on the ldm, so I could potentially wreck a live LDM
if I did that. (or maybe not if it looks for the process ID the
"hard way" ??? )

I talked to Russ about  syncing netCDF file cost.  He seems to think that
the current mechanism is adequate for almost all our users, so I not going
to change it. One is always welcome to make mods for their own site needs
and make changes as you have already done.

As long as the data is actually available after it's written before a
sync I guess that's OK. I was only using ncdump to look at the record
dimension (stations) periodically. MAybe the data is all available and
the sync just updates the dimension variables... I don't know. For the
synoptic decoder especially, for what we were decoding, and the way we were invoking the decoder, the decoder had a very long life
span and it was using the same output file for a very long time. Thus
hours would go by without a sync and the record (stations) value would
not change, and it was just confusing. The actual file size was growing
this whole time of course, so maybe the data really is accessible during
this time.