Parvin, What you are looking for is already available in the "bleeding edge", unreleased version of Solr, which will become version 4.0 sometime later this year. You can download it at [1] and test it out. The feature is called SolrCloud [2] and it replaces the old replication mechanism in 1.x and 3.x versions. Instead of slaves pulling the whole index from masters, the masters will forward individual updates to the slaves. Note that this feature is still under development and certain things will change before 4.0 release, but it is pretty stable and even in use in production some places.
[1] https://builds.apache.org/job/Solr-trunk/lastSuccessfulBuild/artifact/artifacts/ [2] http://wiki.apache.org/solr/SolrCloud -- Jan Høydahl, search solution architect Cominvent AS - www.facebook.com/Cominvent Solr Training - www.solrtraining.com On 21. mai 2012, at 09:25, Parvin Gasimzade wrote: > Hi, > > I am using solr with replication. I have one master that indexes data and > two slaves which pulls index from master and responds to the queries. > > My question is, how can i create fault tolerant architecture? I mean what > should i do when master server crashes? I heard that repeater is used for > this type of architecture. Then, do I have to create one master, one slave > with repeater and one slave? > > Another question is, if master crashes then does slave with repeater start > indexing authomatically or should i configure it manually? > > I asked similar question on the stackoverflow : > http://stackoverflow.com/questions/10597053/fault-tolerant-solr-replication-architecture > > Any help will be appreciated. > > Regards, > Parvin