[GRASS-user] error: "Too many links" with r.in.gdal

[Stefan Luedtke]

Thanks a lot. 

I split the mapsets into single ones for each parameter I am considering
and forwarded the other suggestions to the admin :-)

cheers


On Tue, 2012-05-08 at 21:16 +0200, Markus Neteler wrote: 

> Hi Stefan,
> 
> On Tue, May 8, 2012 at 7:05 PM, Stefan Luedtke <sluedtke at gfz-potsdam.de> wrote:
> > Dear friends,
> >
> > I am running r.in.gdal on a RED HAT system, ext3 file system, grass version
> > is 6.4.1. Am not the admin of the box, so I donno much about it.
> >
> > After importing 4 geotiffs with r.in.gdal, each holds ~7300 layers, I got
> > the error message "Too many links".
> >
> > I checked the thread
> > http://www.mail-archive.com/grass-dev@lists.osgeo.org/msg08547.html and that
> > says something about splitting the mapset. I dont understand all the rest in
> > there. By importing 4 of 5 tiffs I almost got the number of 32000 that was
> > mentioned in that link too,
> 
> The problem is this:
> 
> > 7300 * 4
> [1] 29200
> 
> If you start to add one more (big) multilayer GeoTIFF, you reach the
> magic number
> of 32000 for the ext3 filesystem. You have to know that a GRASS raster
> map is stored in a series of subdirectories. Once you exceed ~ 32000 maps in a
> mapset (which is also a directory effectively), you are out of range for ext3:
> 
> http://en.wikipedia.org/wiki/Ext3#Functionality
> -> "A directory can have at most 31998 subdirectories, because an inode
>      can have at most 32000 links."
> 
> > the following command echos:
> >
> > GRASS 6.4.1
> > (mekong_latlong):~/grassdata/mekong_latlong/era_interim/cell_misc > ls -l|wc
> > -l
> > 31999
> 
> Exactly (it was also my problem in the past as you have seen in the
> previous posting).
> 
> > Any suggestions.
> 
> Sure, several possibilities;
> - convince, if the kernel permits, your admin to update to ext4
> filesystem (I did that
>    already, no big deal)
> - convince, if the kernel permits, your admin to offer a disk
> partition with xfs filesystem
>   ( we are using this here in FEM for our Terabytes of GRASS data)
> - if not possible, simply make 2 (or more) separate mapsets. Thanks to g.mapsets
>    you can still have all maps visible (i.e., in the current map
> search path), so zero
>    efforts when it comes to analysis. In my pre-xfs years, I used to
> split time series
>    by years.
> 
> Nowadays the ext3 mapset is a bit limited when it comes to massive geodata
> processing...
> 
> Hope this helps,
> Markus
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/grass-user/attachments/20120509/c46c2c7b/attachment.html