You could prevent queries to the master by limiting what IP addresses are allowed to communicate with it, or by modifying web.xml to put different security on /update vs. /select .
We took a simplistic approach. We did some load testing, and discovered that we could handle our expected update load and our query load on the master. We use replication just for failover in case the master dies (in which case updates would be held up until the master was fixed). (We also have security on both /update and /select -- just different security). JRJ -----Original Message----- From: Patrick Sauts [mailto:patrick.via...@gmail.com] Sent: Saturday, September 10, 2011 11:26 AM To: solr-user@lucene.apache.org Subject: Re: Master Slave Question Real Time indexing (solr 4) or decrease replication poll and auto commit time. 2011/9/10 Jamie Johnson <jej2...@gmail.com> > Is it appropriate to query the master servers when replicating? I ask > because there could be a case where we index say 50 documents to the > master, they have not yet been replicated and a user asks for page 2, > when they ask for page 2 the request could be sent to a slave and get > 0. Is there a way to avoid this? My thought was to not allow > querying of the master but I'm not sure that this could be configured > in solr >
