You don't need multi-core. Solr already does this automatically. It creates a new Searcher and auto-warms the cache.
But, it will still be slow. If you use auto-warming, it uses most of one CPU, which slows down queries during warming. Also, warming isn't perfect, so queries will be slower after switching to the new Searcher. If you don't use warming, the cold cache will make queries slower. There is no way to get around this. Solr throws away all the caches after replication, so there is a performance hit. In the system I ran, it took a few minutes to recover, so I staggered the replications 10 minutes apart across the search farm. wunder On Apr 9, 2010, at 3:00 AM, Marco Martinez wrote: > Hi Marcin, > > This is because when you do the replication, all the caches are rebuild > cause the index has changed, so the searchs performance decrease. You can > change your architecture to a multicore one to reduce the impact of the > replication. Using two cores, one to do the replication, and other to > search, when the replication is done, do a swap of the cores so the caches > are updated all the time. > > Regards > > > Marco Martínez Bautista > http://www.paradigmatecnologico.com > Avenida de Europa, 26. Ática 5. 3ª Planta > 28224 Pozuelo de Alarcón > Tel.: 91 352 59 42 > > > 2010/4/9 Marcin <[email protected]> > >> Hi guys, >> >> I have noticed that Master/Slave replication process is slowing down slave >> read/search performance during replication being done. >> >> >> please help >> cheers >>
