No, the load balancing is based on random selection of replicas and CPU is not consulted. There are limited ways to influence the replica selection, see https://lucene.apache.org/solr/guide/8_4/distributed-requests.html#shards-preference-parameter
If a replica fails then the query fails and an error is returned. I think (but I am not sure) that SolrJ retries the request on some specific errors in which case a different replica may be selected and the request may succeed. IMO, these are two weak areas of Solr right now. Suggestions/patches are welcome :-) On 12/11/20, Dominique Bejean <dominique.bej...@eolya.fr> wrote: > Hi, > > Is there in Solrcloud any load balancing based on CPU load on Solr nodes ? > > If for shard a replica fails to handle a query, the query is sent to > another replica in order to be completed ? > > Regards > > Dominique > -- Regards, Shalin Shekhar Mangar.
