On 4/17/2015 7:21 AM, Vincenzo D'Amore wrote: > this morning I have optimised my SolrCloud cluster (3 instances). > I have many collections, all are in shard and replica for each node. > At the end of optimisation task (about 10 minutes) all cores are optimised > on every node. > > How can be sure than also reload affects all the cores?
The optimize command is sent at the core level, to a specific machine, but sets in motion an optimize of the entire collection, one core at a time. The optimize update command ignores distrib=false -- it always optimizes the entire collection. If you send a RELOAD action to a core in a collection, it will only affect that core. There is a separate RELOAD action on the Collections API which will reload every core in the collection on all servers. Perhaps we should change how optimize works, and provide an OPTIMIZE action on the Collections API, so it works much the same as RELOAD. I remember seeing an issue in Jira about adding distrib=false support to optimize, but now I can't find it. Changing optimize to work like RELOAD would fix that issue. Thanks, Shawn
