[Mondrian] VFS issue

[Will Gorman]

Hey Guys,

We are still using VFS in Kettle and Mondrian, Kettle 3.2.0 ended up
building from source the latest VFS because they haven't released 2.0 of
VFS yet.  I'm not sure if that would solve these issues or not.

In Pentaho's BI Platform, we access our schema files using a custom VFS
provider so that it can access our solution repository, so we rely on
VFS.

Here is a link to Kettle's custom built VFS 2.0 jar if you want to see
if it resolves the issues:
http://source.pentaho.org/svnkettleroot/Kettle/trunk/libext/commons/commons-vfs-2.0-20090205.jar

Will

On Fri, 2009-07-17 at 15:08 -0400, Matt Campbell wrote:
> I did do a fair amount of searching when we first encountered this
> problem and couldn't find much helpful information at all.  When I get
> a chance I'll see if there is a newer version of VFS.
> 
> On Fri, Jul 17, 2009 at 2:25 PM, Julian Hyde <jhyde at pentaho.com>
> wrote:
>         Has anyone done a google search for the issue you are
>         experiencing? Is there a newer version of VFS that might fix
>         the problem?
>          
>         Will, what was the other technology you used?
>          
>         I'm OK with switching technology as long as it doesn't break
>         existing apps.
>          
>         Julian
>         
>                 
>                 ______________________________________________________
>                 From: mondrian-bounces at pentaho.org
>                 [mailto:mondrian-bounces at pentaho.org] On Behalf Of
>                 Matt Campbell
>                 Sent: Friday, July 17, 2009 8:16 AM
>                 To: Mondrian developer mailing list
>                 Subject: [Mondrian] VFS issue
>                 
>                 
>                 
>                 Awhile back we discovered that Util.readVirtualFile()
>                 can throw exceptions when large numbers of requests
>                 for an HTTP resource happen.  Awhile back I added a
>                 call to file.close() at the end of that method, which
>                 had seemed to make things a bit better in testing.
>                 I'm still seeing socketConnect exceptions when
>                 requesting more than a few hundred http requests.  In
>                 practice this can happen when we have as few as 15
>                 concurrent users running reports.
>                 
>                 There appears to be no such problem with the pre-VFS
>                 method that Mondrian used for loading a URL.  I
>                 created a Unit test which loads 100,000 pages via the
>                 old Util.readURL method.  That runs fine with no
>                 exceptions.  The VFS method fails with fewer than 1000
>                 requests.
>                 
>                 Would anyone object to reverting back to the old
>                 Util.readURL for http requests?  Alternatively, I
>                 think Will Gorman said at one point that they
>                 abandoned VFS for other Pentaho projects in favor of
>                 some other library--would that be worth considering
>                 for Mondrian?
>                 
>                 
>                 
>         
>         _______________________________________________
>         Mondrian mailing list
>         Mondrian at pentaho.org
>         http://lists.pentaho.org/mailman/listinfo/mondrian
>         
> 
> _______________________________________________
> Mondrian mailing list
> Mondrian at pentaho.org
> http://lists.pentaho.org/mailman/listinfo/mondrian