Re: Solr7.4 core node and shard replica numbering question

[Shawn Heisey <ap...@elyograg.org>]

On 8/29/2018 6:00 AM, Synet Spambox wrote:
> The core numbers and the replica numbers are up to 10 (double the amount of servers)?
> It is confusing for me...
> Does it matter?
> Do I miss something ?

Something changed somewhere regarding core naming in SolrCloud, and it 
doesn't behave in the same way as it used to.

The names don't actually matter -- SolrCloud uses the info in ZooKeeper 
to locate everything, and that info will match the actual names that are 
created.  If it bothers you enough, the source code is available to 
anyone and you're welcome to figure it out and submit a patch to fix it.

I'm curious why you have increased jute.maxbuffer to such a large 
value.  Solr 5.0 and later will generally only create large info in ZK 
at Solr service restart, and only if you have a large number of 
collections or shard replicas.  The overseer queue can get very large in 
these situations, and you can run into this problem with ZK:

https://issues.apache.org/jira/browse/ZOOKEEPER-1162

Thanks,
Shawn