[Building Sakai] Anybody using search in a clustered setting ?

[Joshua Swink]

At UC Merced we had the same problem, which turned out to be caused by
setting localIndexBase to a shared directory, resulting in the
different instances interfering with each others' index files.

Currently we have two dedicated indexer instances which are not part
of the customer-serving pool. One of these serves as the search
server, and it's working out fine.

The configuration for the indexers (one of which performs searching as well) is:

search.indexbuild=true
# This directory is not shared
localIndexBase at org.sakaiproject.search.api.JournalSettings=${sakai.home}/indexwork
# This directory is shared between the indexers via NFS
sharedJournalBase at org.sakaiproject.search.api.JournalSettings=${sakai.home}/searchjournal
searchServer at org.sakaiproject.search.api.SearchService=true
sharedKey at org.sakaiproject.search.api.SearchService=XXXXXXXXXXXXXXXXXXXXXXXXXXX
# Set this to TRUE for a single indexing server, or FALSE when there
are multiple indexing servers.
localSegmentsOnly at org.sakaiproject.search.api.SearchService.SegmentStore=false

The configuration for the client instances is:

search.indexbuild=false
searchServerUrl at org.sakaiproject.search.api.SearchService=<SEARCH_SERVER_HTTP_URL>/sakai-search-tool/xmlsearch/
sharedKey at org.sakaiproject.search.api.SearchService=XXXXXXXXXXXXXXXXXXXXXXXXXXX

--
Joshua Swink

On Fri, Nov 5, 2010 at 9:29 AM, George Pipkin <gpp8p at virginia.edu> wrote:
> Hi Everybody -
>
>    A year or so ago we decided to turn off the search feature because
> at startup,
> more often than not Lucene detected a bad index and decided to start
> re-indexing everything. This created intolerable load levels, and we had
> to turn it off.  However, we'd like to get search going again, and one
> option
> we're looking at is the one being used at Cambridge where 'good' indexes are
> copied from other nodes in the cluster.
>
>    I was wondering if anybody else is running search in a clustered
> setting, and if
> so, how it's being supported. Is anybody other than Cambridge using the
> scripts
> written by Daniel Parry that synchronize indexes ?  Have they been
> modified ?  Are
> there other approaches to this problem ?
>
>                                         - George Pipkin
>                                            U.Va.
> _______________________________________________
> sakai-dev mailing list
> sakai-dev at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/sakai-dev
>
> TO UNSUBSCRIBE: send email to sakai-dev-unsubscribe at collab.sakaiproject.org with a subject of "unsubscribe"
>