First, do not think in terms of cores, think "replicas" ;). And do not, use the "core admin" bits of the admin UI to do any SolrCloud-related operations. It's possible, but far too easy to get wrong.
Use the collections API instead. Second, 600 collections, assuming all on a single cluster is getting pretty big. Solr will start to have issues as the number of collections increases much more. But to your point: The simplest thing to do is use the ADDREPLICA and DELETEREPLICA commands from the collections API and be done with it. For ADDREPLICA, you can specify which node it is supposed to appear on. So let's say one of your nodes is just gone that contained a replica for collection1, shard1. Hardware fault or similar, and it's not coming back. You bring up a new Solr node to replace it and: 1> issue an ADDREPLICA for collection1, shard1 that specifies the new node. Just wait for a while and it'll replicate the necessary data from the leader and start serving queries. You do not have to stop indexing during this process, although you might want to. Nor do you have to stop querying. 2> issue a DELETEREPLICA for the old replica. That'll clear it out of the Zookeeper cluster state. The order of these streps is irrelevant. bq: or if accidentally i delete the data dir You shouldn't have to do anything except restart the node. It should replicate the index from the leader. That said, you shouldn't have to manipulate the directories under SOLR_HOME manually, if you find yourself doing this regularly it's a red flag.. There's no single command to say "add a replica on nodeX for every replica on nodeY" though. So you'd have to do this for each replica on the disappeared node. All the necessary information is present in the clusterstate, and such a command is possible, but nobody's done it yet. It'd be possible to create an external utility to do this too. Best, Erick On Tue, Aug 18, 2015 at 10:07 AM, Olivier Damiot <olivier.dam...@gmail.com> wrote: > hello, > > i'am a bit confused about how solr cloud recovery is supposed to work > exactly in the case of loosing a single node completely. > > My 600 collections are created with > numShards=3&replicationFactor=3&maxShardsPerNode=3 > > However, how do i configure a new node to take the place of the dead > node, or if accidentally i delete the data dir ? > > I bring up a new node which is completely empty (empty data dir), > install solr, and connect it to zookeeper.Is it supposed to work > automatically from there? All my shards/replicas on this node as down > (i suppose because there is no cores in data dir). > > Do I need to recreate the cores first? > > Can i copy/paste data directory from another node to this one ? I > think no because i should rename all variables in core.properties > which are dedicated for each node (like name or coreNodeName) > > thanks, > > Olivier Damiot
