On 12/6/2011 1:01 AM, Husain, Yavar wrote:
In solrconfig.xml I was experimenting with Indexing Performance. When I set the
maxDocs (in autoCommit) to say 10000 documents the index size is double to if I
just dont use autoCommit (i.e. keep it commented, i.e commit at the end only
after adding documents).
Does autoCommit affect the index size and if so why (as it should only affect
performance) ?
In other messages you have mentioned using MSSQL as a data source, so I
am taking a wild guess and assuming that you are running on Windows. If
that's not the case, then I don't know why you are seeing this behavior.
On Windows, the timing involved when index segments are merged results
in the old index segments being left on the disk, because they are still
open for reading when Solr tries to delete them, and on Windows you
cannot delete an open file. Later merge/optimize passes will remove
files left over from the previous passes, but will themselves suffer
from the same problem.
I am guessing that when you don't have autoCommit turned on, Solr never
opens an IndexReader on the segments as they are created, so Windows has
no trouble deleting old segments after merging/optimizing. With
autoCommit turned on, Solr opens the index segments and makes them
searchable after every commit.
Thanks,
Shawn
