On Tue, Aug 18, 2015, at 06:35 PM, Erick Erickson wrote: > As SolrCloud works its way into larger and larger situations, the > possibility of a node just failing (hardware fault, somebody spills > coffee on it, whatever) increases. Currently to replace the whole > node, the best I can come up with is ADDREPLICA/DELETEREPLICA for all > the replicas on the node. > > All the information should be in ZK to allow a command saying > something like "for every replica on nodeX, make a new one on nodeY". > True, the names would be different.... > > Worth a JIRA? It could be a Collections API call or a SolrJ utility or > even a different script. Or maybe just a utility that would generate > the appropriate URLs for someone to submit themselves. > > I'm not sure there's enough need but thought I'd see what others > thought before raising a JIRA.
I like the idea. If it were an API, we could then expose it via that UI to allow a "known bad" node to be decommissioned easily, by looking at the cloud graph. Upayavira --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
