bq: ...in the sense that there's only one canonical copy. Agreed, and as you say that copy is kept in ZooKeeper.....
And I pretty much guarantee that the internal solrconfig object is NOT shared. I doubt the schema object is shared, but it seems like it could be with some work. But the savings potential here is rather small unless you have a large number of cores. The LotsOfCores option is really, at this point, orthogonal to SolrCloud, I don't think (and we have some anecdotal evidence) that they don't play nice together.... Erick On Wed, Oct 9, 2013 at 12:17 PM, Shawn Heisey <s...@elyograg.org> wrote: > On 10/9/2013 6:24 AM, Erick Erickson wrote: >> >> Hmmm, I hadn't thought about that before. The shareSchema >> stuff is keyed off the absolute directory (and timestamp) of >> the schema.xml file associated with a core and is about >> sharing the internal object that holds the parsed schema. >> >> Do you know for sure if the fact that this is coming from ZK >> actually shares the schema object? 'Cause I've never >> looked to see and it would be a good thing to have in my >> head... > > > With SolrCloud, I have no idea whether the actual internal objects are > shared. Just now I tried to figure that out from the code, but I don't > already have an understanding of how that code works, and a quick glance > isn't enough to gain that knowledge.I can guarantee that you have a much > deeper understanding of those internals than I do! > > My comments were to indicate that SolrCloud creates a situation where the > config/schema are shared in the sense that there's only one canonical copy. > > Thanks, > Shawn >
