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

[netCDF #XIX-708449]: Memory leaks using netcdf4

This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.

  • Subject: [netCDF #XIX-708449]: Memory leaks using netcdf4
  • Date: Tue, 22 Mar 2011 09:00:54 -0600
OK, I will close this ticket.  You can lessen the amount of memory
used in the chunk cache of the HDF5 layer at netCDF configuration
time by changing values of these options to the configure script:

  --with-default-chunks-in-cache=<integer>
                          Specify the number of chunks to store in default
                          per-variable cache.
  --with-chunk-cache-size=<integer>
                          Specify default file cache chunk size for HDF5 files
                          in bytes.
  --with-chunk-cache-nelems=<integer>
                          Specify default maximum number of elements in the
                          file chunk cache chunk for HDF5 files (should be
                          prime number).
  --with-chunk-cache-preemption=<float between 0 and 1 inclusive>
                          Specify default file chunk cache preemption policy
                          for HDF5 files (a number between 0 and 1,
                          inclusive).

If you see any more evidence of memory leaks, please let us know.  Thanks.

--Russ

> Le 21/03/2011 18:18, Unidata netCDF Support a écrit :
> > Ludovic
> >
> >> I have download 4.1.2-rc1 on a new pc and I don't have problem with
> >> nc_create.
> >>
> >> I run the tst_mem_leak and the memory consumption is increasing.
> >> I see it with top or ksysguard.
> >> Can you check with ksysguard and tell me if you see the same.
> > I see a slow growth in memory use with top or the Fedora System
> > Monitor application, but it stops growing after it has about 20
> > MBytes.  Changing the usleep(10000) call to usleep(100) makes it reach
> > 20 MB much faster, and then the file continues to grow (I let it
> > increase to over 500 MB) but there is no more growth in memory use.
> >
> > So I can't reproduce any memory leaks with your demo program, using
> > valgrind, top, or the Fedora System monitor.
> >
> > --Russ
> >
> > Russ Rew                                         UCAR Unidata Program
> > address@hidden                      http://www.unidata.ucar.edu
> >
> >
> >
> > Ticket Details
> > ===================
> > Ticket ID: XIX-708449
> > Department: Support netCDF
> > Priority: Normal
> > Status: Closed
> >
> Thank you: I have the same behaviour.
> The memory in use is growing (about each 2 seconds) to reach 23MB
> 
> Then, I've change the test to add a nc_sync each 10 seconds and the
> memory in use is growing each nc_sync to reach 21MB.
> 
> Best regards,
> Ludovic
> 
> 

Russ Rew                                         UCAR Unidata Program
address@hidden                      http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: XIX-708449
Department: Support netCDF
Priority: Normal
Status: Closed