I see your point. In the custom routing case where it _is_ reasonable to delete all replicas in a shard, DELETESHARD does the trick.
It's reasonable to raise a JIRA I think so we have a record of the discussion/decision. Erick On Sun, Nov 8, 2015 at 9:13 AM, Yago Riveiro <yago.rive...@gmail.com> wrote: > I don't know if this behaviour makes sense but in my IMHO the last replica of > a shard should be removed only by the DELETESHARD command ... > > If you don't not notice about this behaviour, the result is the deletion of > the whole shard, and as a consequence, the deletion of the shard in the > clusterstate.json ... with all the pain associated. > > > > ----- > Best regards > -- > View this message in context: > http://lucene.472066.n3.nabble.com/DELETEREPLICA-command-shouldn-t-delete-de-last-replica-of-a-shard-tp4239054.html > Sent from the Solr - User mailing list archive at Nabble.com.
