Ronald: Actually, people generally don't search on master ;). The idea is that master is configured for heavy indexing and then people search on the slaves which are configured for heavy query loads (e.g. memory, autowarming, whatever may be different). Which is it's own problem since the time the slaves poll won't necessarily be the exact same wall-clock time.....
SolrCloud doesn't use replication except in certain recovery scenarios. In normal operations, documents are forwarded to each replica and indexed separately on all nodes. That's about the only way to support Near Real Time. Best, Erick On Tue, Jul 25, 2017 at 9:39 AM, Stanonik, Ronald <rstano...@ucsd.edu> wrote: > Bingo! Right on both counts! opensearcher was false. When I changed it to > true, then I could see that master(searching) and master(replicable) both > changed. And autocommit.maxtime is causing a commit on the master. > > Who uses master(replicable)? It seems for my simple master/slave > configuration master(searching) is the relevant version. Maybe solr cloud > uses master(replicable)? > > Thanks, > > Ron > >
