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

Re: fstat update



Thanks for the response; since we can't have the program creating the data on 
the same machine as the IDL machine I think we might try having the "data 
program" write the data file onto the IDL machine's hard drive, via the 
network.
I think, however, that the problem isn't soley one of NFS (or equiv) update 
frequency since fstat reported the same file size every fews second for the 
whole 8 hour run!

Thanks again,

Justin

In article <88iuduugyk.fsf@catspaw.jpl.nasa.gov>, 
Vapuser<vapuser@catspaw.jpl.nasa.gov> wrote:
>ashmall@my-dejanews.com (Justin Ashmall) writes:
>
>> Dear All,
>> Has anyone had a problem like this?
>
>> We have a widget prog that periodically checks a file for new data,
>> it does this by monitoring the file size as returned by fstat. This
>> usually works fine but we have found on some (network) drives fstat
>> always returns the same value, namely the file size it first saw,
>> even though the file is growing.  The only fix we've found is to
>> constantly close and re-open the file which is a rather inelegant,
>> if not inefficient, solution.  It doesn't appear to be a buffering
>> problem since we can see the file size changing (across the network)
>> using the command line or a file manager.
>
>> Any suggestions?
>> (We're using IDL 5.1.1 under NT4)
>> 
>> Many thanks,
>> 
>> Justin
>
>  I've seen this problem in a UNIX environment when running an IDL
>  routine on one machine that was querying the size of a file which
>  resided on a disk that was NFS mounted from another machine. I am
>  unfamiliar with NT, so I don't know if 'network disk' is a fair
>  analog of NFS. In this case it was a feature of the operating
>  system/NFS, the file size didn't change as frequently on the NFS
>  client, therefore fstat reported EOF correctly as far as it could
>  see. The work around was to run the program on the machine which
>  hosted (i.e. for which the disk was local) the disk containing the
>  file in question.
>
>  Hope this helps in some small way.
>
>whd
>