+1 on making it consistent in the docs for now. The code is a mess of terminology, indeed … we use both shards and slices, both replicas and cores, and most confusingly String core names and String coreNode names, etc
Cheers, Andrzej On 2026/10/15 12:14:24 Gus Heck wrote: > Works for me, > > Can we also agree to always call the things that we combine into an index > (user-managed term) / collection (cloud term) shards and all the copies of > those parts replicas (and never "core")? Also, since all replicas are > equal (until one is identified/elected as the leader of a shard), the > minimum number of replicas is 1 (never 0) > > I'm only suggesting this for documentation for now. Changing the code base > naming conventions would be a whole other (much bigger) monster of course. > > The word "Core" only makes sense in the case where you have single stand > alone index with no shards. In that case, it could be thought of as the > "central thing". In all other configurations, "core" conjures (very wrong) > notions of centrality but it describes the smallest leaf node unit with the > highest numerical count across the system. Learning to ignore the English > meanings associated with the word "core" was one of the most confusing > parts of learning Solr when I first started working with it. > > The CPU-related usage of the word core is the only usage that even sort of > justifies its use, but even with CPUs we are talking about physically > co-located, permanent, on-chip segments of a single *central* processing > unit, unlike our usage which involves things scattered across racks or even > data centers that may come and go with varying configurations, so that > analogy is weak at best. > > -Gus > > On Wed, Oct 14, 2026 at 7:18 PM David Smiley <[email protected]> wrote: > > > I wish to seek a nomenclature decision. I make the following assertions: > > > > (A) a Solr *node* has two modes: > > * SolrCloud > > * standalone > > > > (B) a Solr *cluster* has two types: > > * SolrCloud > > * user managed > > > > Thus the words "standalone" and "user managed" have their place depending > > on context. There is too much precedent for the word "standalone" to > > ignore it, and it's a fine word for the context in which I propose it > > here. On the other hand, it's not a suitable word to characterize a > > cluster, and hence we have the term "user managed" that Cassandra > > championed. > > > > Consequence (just listing one for now): > > * When you start a Solr node, the flag "--user-managed" shall be renamed > > to "--standalone". > > Related PR under review now: https://github.com/apache/solr/pull/5121 > > > > Perhaps Eric might argue for "--solrcloud-enabled=false" as he has sought > > standardizing on all binary decisions that could be expressed with an > > "enabled" suffix to do so... albeit I think in this case, it doesn't look > > as nice to me. But I don't mind. > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [email protected] > > For additional commands, e-mail: [email protected] > > > > > > -- > http://www.needhamsoftware.com (work) > https://a.co/d/b2sZLD9 (my fantasy fiction book) > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
