jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Lorenzo Dini <Lorenzo.D...@cern.ch>
Subject Re: Migrating to DataStore - large memory usage of importxml
Date Tue, 15 Apr 2008 14:04:43 GMT

I have 100GB stored in binaries and for me is impossible to do a system view
with binaries included.

What I do is:

Source installation:
 - export a system view excluding binaries

Destination installation:
 - setting a binary location to a fake location in the configuration
 - importing the system view (that generates 0 length files in the fake
binary location)
 - copying the real files from the source to the destination real binary
 - changing back the location to point the just copied files
 - if necessary, refresh the indexes

Don't know if it helps but for me it works.


On 15/04/2008 15:31, "Ivan Latysh" <ivanlatysh@gmail.com> wrote:

> sbarriba wrote:
>> Firstly worth saying how good the DataStore solution is - well done to the
>> team.
>> We're considering migrating some pretty large repositories over to a
>> DataStore implementation but are limited by the memory usage of the
>> importxml command. We can run exportSysView to produce a 3GB XML backup no
>> problem.
>> Any suggestions on how to re-import an XML backup of 3GB+ on a machine with
>> 1 or 2 GB? Incremental export/import, e.g. folder by folder, is tricky as
>> the workspaces have various internal cross-references.
> JCR-Backup address this issue, by using streaming API only.
> We didn't test on 3G import, but hundreds of megs are going in and out fine.
> Also nt:file nodes are backed up in native form, no BASE64 encoding, that
> significantly reduce backup size and time.

View raw message