Hi Erick and Shawn

   Thanks a lot


2013/9/17 Erick Erickson <erickerick...@gmail.com>

> Here's a blog about tlogs and commits:
>
> http://searchhub.org/2013/08/23/understanding-transaction-logs-softcommit-and-commit-in-sorlcloud/
>
> And here's Mike's excellent segment merging blog
>
> http://blog.mikemccandless.com/2011/02/visualizing-lucenes-segment-merges.html
>
> Best,
> Erick
>
>
> On Tue, Sep 17, 2013 at 6:36 AM, Shawn Heisey <s...@elyograg.org> wrote:
>
> > On 9/17/2013 12:32 AM, YouPeng Yang wrote:
> > > Hi
> > >    Another werid problem.
> > >    When we setup the autocommit properties, we  suppose that the index
> > > fille will created every commited.So that the size of the index files
> > will
> > > be large enough. We do not want to keep too many small files as [1].
> > >
> > >    How to control the size of the index files.
> >
> > An index segment gets created after every hard commit.   In the listing
> > that you sent, all the files starting with _28w are a single segment.
> > All the files starting with _28x are another segment.
> >
> > Solr should be merging the segments when you get enough of them, unless
> > you have incorrectly set up your merge policy.  The default number of
> > segments that get merged is ten.  When you get ten segments, they will
> > be merged down to one.  This repeats until you have ten merged segments.
> >  At that point, those ten merged segments will be merged to make an even
> > larger segment.
> >
> > You can bump up the number of open files allowed by your operating
> > system.  On Linux, this is controlled by the /etc/security/limits.conf
> > file.  Here are some example config lines for that file:
> >
> > elyograg        hard    nofile  6144
> > elyograg        soft    nofile  4096
> > root            hard    nofile  6144
> > root            soft    nofile  4096
> >
> > Alternatively, you can reduce the required number of files if you turn
> > on the UseCompoundFile setting, which is in the IndexConfig section.
> > This causes Solr to create a single file per index segment instead of
> > several files per segment.  The compound file may be slightly less
> > efficient, but the difference is likely to be very small.
> >
> >
> https://cwiki.apache.org/confluence/display/solr/IndexConfig+in+SolrConfig
> >
> >
> >
>

Reply via email to