On 06/19/2013 03:21 AM, Otis Gospodnetic wrote: > You could call the optimize command directly on slaves, but specify > the target number of segments, e.g. > ..../solr/update?optimize=true&maxSegments=10 > > Not sure I recommend doing this on slaves, but you could - maybe you > have spare capacity. You may also want to consider not doing it on > all your slaves at the same time...
IIUC this assumes your slaves do not replicate too often, otherwise replication would reset the index to whatever number of segments the master has. You could still perform an optimize with maxSegments after every replication, if it's acceptable in the situation you are in. However, if you need slaves to update every 2-5 minutes, that would be impractical and wasteful. Is this correct? If so, how to find a fair compromise/balance between master and slave merge factors if you need very frequent indexing of new documents (say continuous) on the master and up-to-date indexes on the slaves (say 2-5' pollInterval)? -- Cosimo
